List of usage examples for java.lang Math sqrt
@HotSpotIntrinsicCandidate public static double sqrt(double a)
From source file:ch.epfl.lsir.xin.test.ItemAverageTest.java
/** * @param args//from w w w .java 2 s. co m */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//ItemAverage"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File(".//conf//ItemAverage.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); DataLoaderFile loader = new DataLoaderFile(".//data//MoveLens100k.txt"); loader.readSimple(); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; int F = 5; logger.println(F + "- folder cross validation."); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { logger.println("Folder: " + folder); logger.flush(); System.out.println("Folder: " + folder); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = new HashMap<String, Integer>(); HashMap<String, Integer> itemIDIndexMapping = new HashMap<String, Integer>(); for (int i = 0; i < dataset.getUserIDs().size(); i++) { userIDIndexMapping.put(dataset.getUserIDs().get(i), i); } for (int i = 0; i < dataset.getItemIDs().size(); i++) { itemIDIndexMapping.put(dataset.getItemIDs().get(i), i); } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } trainRatingMatrix.calculateGlobalAverage(); RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } System.out.println("Training: " + trainRatingMatrix.getTotalRatingNumber() + " vs Test: " + testRatingMatrix.getTotalRatingNumber()); logger.println("Initialize a recommendation model based on item average method."); ItemAverage algo = new ItemAverage(trainRatingMatrix); algo.setLogger(logger); algo.build(); algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); logger.flush(); System.out.println(trainRatings.size() + " vs. " + testRatings.size()); double RMSE = 0; double MAE = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID())); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " MAE: " + MAE + " RMSE: " + RMSE); logger.flush(); // System.out.println("MAE: " + MAE + " RMSE: " + RMSE); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Final results: MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); logger.flush(); //MAE: 0.8173633324758338 RMSE: 1.0251973503888645 (MovieLens 100K) }
From source file:kishida.cnn.NeuralNetwork.java
public static void main(String[] args) throws IOException { NeuralNetwork nn = new NeuralNetwork(); nn.getLayers()/* w w w .j a v a 2s .c o m*/ .addAll(Arrays.asList(new InputLayer(20, 20), new ConvolutionLayer("conv1", 3, 7, 2, 1, true), new MaxPoolingLayer("pool", 3, 2), new MultiNormalizeLayer("norm1", 5, .0001f, true), new FullyConnect("test", 3, 0, 1, new LogisticFunction(), true))); nn.init(); nn.random.nextInt(); StringWriter sw = new StringWriter(); nn.writeAsJson(sw); System.out.println(sw); // ?????????????? StringReader sr0 = new StringReader(sw.toString()); NeuralNetwork nn0 = nn.readFromJson(sr0); nn0.init(); ConvolutionLayer conv1o = (ConvolutionLayer) nn.findLayerByName("conv1").get(); ConvolutionLayer conv1r = (ConvolutionLayer) nn0.findLayerByName("conv1").get(); System.out.println("org:" + Arrays.toString(conv1o.getFilter())); System.out.println("red:" + Arrays.toString(conv1r.getFilter())); double loss = IntStream.range(0, conv1o.getFilter().length) .mapToDouble(i -> (conv1o.getFilter()[i] - conv1r.getFilter()[i]) * (conv1o.getFilter()[i] - conv1r.getFilter()[i])) .sum(); System.out.println(Math.sqrt(loss)); NeuralNetwork v = NeuralNetwork.readFromJson(new StringReader("{\n" + " \"weightDecay\" : 5.0E-4,\n" + " \"miniBatch\" : 128,\n" + " \"random\" : \"c3EAfgAAAT/wWGBKFyCXAAATnQ6sF654\",\n" + " \"imageRandom\" : \"c3EAfgAAAAAAAAAAAAAAAAAABd7s70R4\",\n" + " \"momentam\" : 0.9,\n" + " \"layers\" : [ {\n" + " \"InputLayer\" : {\n" + " \"width\" : 250,\n" + " \"height\" : 220,\n" + " \"name\" : \"input\"\n" + " }\n" + " }, {\n" + " \"ConvolutionLayer\" : {\n" + " \"name\" : \"conv1\",\n" + " \"filter\" : null,\n" + " \"bias\" : [ 1.0, 1.0, 1.0 ],\n" + " \"filterDelta\" : null,\n" + " \"biasDelta\" : [ 0.0, 0.0, 0.0 ],\n" + " \"stride\" : 2,\n" + " \"filterSize\" : 7,\n" + " \"useGpu\" : true\n" + " }\n" + " }, {\n" + " \"MaxPoolingLayer\" : {\n" + " \"name\" : \"pool\",\n" + " \"size\" : 3,\n" + " \"stride\" : 2\n" + " }\n" + " }, {\n" + " \"MultiNormalizeLayer\" : {\n" + " \"name\" : \"norm1\",\n" + " \"size\" : 5,\n" + " \"threshold\" : 1.0E-4,\n" + " \"useGpu\" : true\n" + " }\n" + " }, {\n" + " \"FullyConnect\" : {\n" + " \"name\" : \"test\",\n" + " \"outputSize\" : 3,\n" + " \"weight\" : [ 0.0014115907, 0.0043465886, 0.01138472, -0.0013297468, " + "-0.0060525155, -0.0109255025, -0.015493984, 0.011872963, -0.0015145391 ],\n" + " \"initBias\" : 0.5, " + " \"bias\" : [ 0.0, 0.2, 0.4 ],\n" + " \"weightDelta\" : [ 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0 ],\n" + " \"biasDelta\" : [ 0.0, 0.0, 0.0 ],\n" + " \"dropoutRate\" : 1.0,\n" + " \"activation\" : \"LogisticFunction\",\n" + " \"useGpu\" : true\n" + " }\n" + " } ],\n" + " \"learningRate\" : 0.01\n" + "}")); System.out.println(nn.random.nextInt()); System.out.println(v.random.nextInt()); v.findLayerByName("test").ifPresent(layer -> { FullyConnect f = (FullyConnect) layer; System.out.println(f.getActivation().getClass()); System.out.println(Arrays.toString(f.getBias())); }); v.init(); v.findLayerByName("test").ifPresent(layer -> { FullyConnect f = (FullyConnect) layer; System.out.println(f.getActivation().getClass()); System.out.println(Arrays.toString(f.getBias())); }); }
From source file:ch.epfl.lsir.xin.test.ItemBasedCFTest.java
/** * @param args//www .j a va 2s .c om */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//ItemBasedCF"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File(".//conf//ItemBasedCF.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); DataLoaderFile loader = new DataLoaderFile(".//data//MoveLens100k.txt"); loader.readSimple(); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; double totalPrecision = 0; double totalRecall = 0; double totalMAP = 0; double totalNDCG = 0; double totalMRR = 0; double totalAUC = 0; int F = 5; logger.println(F + "- folder cross validation."); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { logger.println("Folder: " + folder); System.out.println("Folder: " + folder); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = new HashMap<String, Integer>(); HashMap<String, Integer> itemIDIndexMapping = new HashMap<String, Integer>(); for (int i = 0; i < dataset.getUserIDs().size(); i++) { userIDIndexMapping.put(dataset.getUserIDs().get(i), i); } for (int i = 0; i < dataset.getItemIDs().size(); i++) { itemIDIndexMapping.put(dataset.getItemIDs().get(i), i); } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } trainRatingMatrix.calculateGlobalAverage(); trainRatingMatrix.calculateItemsMean(); RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } System.out.println("Training: " + trainRatingMatrix.getTotalRatingNumber() + " vs Test: " + testRatingMatrix.getTotalRatingNumber()); logger.println("Initialize a item based collaborative filtering recommendation model."); ItemBasedCF algo = new ItemBasedCF(trainRatingMatrix); algo.setLogger(logger); algo.build();//if read local model, no need to build the model algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); logger.flush(); //rating prediction accuracy double RMSE = 0; double MAE = 0; double precision = 0; double recall = 0; double map = 0; double ndcg = 0; double mrr = 0; double auc = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID()), false); if (prediction > algo.getMaxRating()) prediction = algo.getMaxRating(); if (prediction < algo.getMinRating()) prediction = algo.getMinRating(); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; System.out.println("Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Folder --- MAE: " + MAE + " RMSE: " + RMSE); //ranking accuracy if (algo.getTopN() > 0) { HashMap<Integer, ArrayList<ResultUnit>> results = new HashMap<Integer, ArrayList<ResultUnit>>(); for (int i = 0; i < trainRatingMatrix.getRow(); i++) { // ArrayList<ResultUnit> rec = algo.getRecommendationList(i); // results.put(i, rec); ArrayList<ResultUnit> rec = algo.getRecommendationList(i); if (rec == null) continue; int total = testRatingMatrix.getUserRatingNumber(i); if (total == 0)//this user is ignored continue; results.put(i, rec); } RankResultGenerator generator = new RankResultGenerator(results, algo.getTopN(), testRatingMatrix, trainRatingMatrix); precision = generator.getPrecisionN(); totalPrecision = totalPrecision + precision; recall = generator.getRecallN(); totalRecall = totalRecall + recall; map = generator.getMAPN(); totalMAP = totalMAP + map; ndcg = generator.getNDCGN(); totalNDCG = totalNDCG + ndcg; mrr = generator.getMRRN(); totalMRR = totalMRR + mrr; auc = generator.getAUC(); totalAUC = totalAUC + auc; System.out.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); logger.append("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc + "\n"); } } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); System.out.println("Precision@N: " + totalPrecision / F); System.out.println("Recall@N: " + totalRecall / F); System.out.println("MAP@N: " + totalMAP / F); System.out.println("MRR@N: " + totalMRR / F); System.out.println("NDCG@N: " + totalNDCG / F); System.out.println("AUC@N: " + totalAUC / F); System.out.println("similarity: " + config.getString("SIMILARITY")); //MAE: 0.7227232762922241 RMSE: 0.9225576790122603 (MovieLens 100K, shrinkage 2500, neighbor size 40, PCC) //MAE: 0.7250636319353241 RMSE: 0.9242305485411567 (MovieLens 100K, shrinkage 25, neighbor size 40, PCC) //MAE: 0.7477213243604459 RMSE: 0.9512195004171138 (MovieLens 100K, shrinkage 2500, neighbor size 40, COSINE) logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + "\n" + "MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F + "\n" + "Precision@N: " + totalPrecision / F + "\n" + "Recall@N: " + totalRecall / F + "\n" + "MAP@N: " + totalMAP / F + "\n" + "MRR@N: " + totalMRR / F + "\n" + "NDCG@N: " + totalNDCG / F + "\n" + "AUC@N: " + totalAUC / F); logger.flush(); logger.close(); }
From source file:ch.epfl.lsir.xin.test.SVDPPTest.java
/** * @param args//from ww w. j a v a2 s. c o m */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//SVDPP"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File("conf//SVDPlusPlus.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); logger.flush(); DataLoaderFile loader = new DataLoaderFile(".//data//MoveLens100k.txt"); loader.readSimple(); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; double totalPrecision = 0; double totalRecall = 0; double totalMAP = 0; double totalNDCG = 0; double totalMRR = 0; double totalAUC = 0; int F = 5; logger.println(F + "- folder cross validation."); logger.flush(); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { System.out.println("Folder: " + folder); logger.println("Folder: " + folder); logger.flush(); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = new HashMap<String, Integer>(); HashMap<String, Integer> itemIDIndexMapping = new HashMap<String, Integer>(); for (int i = 0; i < dataset.getUserIDs().size(); i++) { userIDIndexMapping.put(dataset.getUserIDs().get(i), i); } for (int i = 0; i < dataset.getItemIDs().size(); i++) { itemIDIndexMapping.put(dataset.getItemIDs().get(i), i); } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { if (testRatings.get(i).getValue() < 5) continue; testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } System.out.println("Training: " + trainRatingMatrix.getTotalRatingNumber() + " vs Test: " + testRatingMatrix.getTotalRatingNumber()); logger.println("Initialize a SVD++ recommendation model."); logger.flush(); SVDPlusPlus algo = new SVDPlusPlus(trainRatingMatrix, false, ".//localModels//" + config.getString("NAME")); algo.setLogger(logger); algo.build(); algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); logger.flush(); //rating prediction accuracy double RMSE = 0; double MAE = 0; double precision = 0; double recall = 0; double map = 0; double ndcg = 0; double mrr = 0; double auc = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID()), false); if (prediction > algo.getMaxRating()) prediction = algo.getMaxRating(); if (prediction < algo.getMinRating()) prediction = algo.getMinRating(); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; System.out.println("Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Folder --- MAE: " + MAE + " RMSE: " + RMSE); //ranking accuracy if (algo.getTopN() > 0) { HashMap<Integer, ArrayList<ResultUnit>> results = new HashMap<Integer, ArrayList<ResultUnit>>(); for (int i = 0; i < trainRatingMatrix.getRow(); i++) { ArrayList<ResultUnit> rec = algo.getRecommendationList(i); if (rec == null) continue; int total = testRatingMatrix.getUserRatingNumber(i); if (total == 0)//this user is ignored continue; results.put(i, rec); } RankResultGenerator generator = new RankResultGenerator(results, algo.getTopN(), testRatingMatrix); precision = generator.getPrecisionN(); totalPrecision = totalPrecision + precision; recall = generator.getRecallN(); totalRecall = totalRecall + recall; map = generator.getMAPN(); totalMAP = totalMAP + map; ndcg = generator.getNDCGN(); totalNDCG = totalNDCG + ndcg; mrr = generator.getMRRN(); totalMRR = totalMRR + mrr; auc = generator.getAUC(); totalAUC = totalAUC + auc; System.out.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); logger.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); } logger.flush(); } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); System.out.println("Precision@N: " + totalPrecision / F); System.out.println("Recall@N: " + totalRecall / F); System.out.println("MAP@N: " + totalMAP / F); System.out.println("MRR@N: " + totalMRR / F); System.out.println("NDCG@N: " + totalNDCG / F); System.out.println("AUC@N: " + totalAUC / F); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + "\n" + "MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F + "\n" + "Precision@N: " + totalPrecision / F + "\n" + "Recall@N: " + totalRecall / F + "\n" + "MAP@N: " + totalMAP / F + "\n" + "MRR@N: " + totalMRR / F + "\n" + "NDCG@N: " + totalNDCG / F + "\n" + "AUC@N: " + totalAUC / F); logger.flush(); logger.close(); }
From source file:ch.epfl.lsir.xin.test.BiasedMFTest.java
/** * @param args/*from w w w .j ava 2 s . c om*/ */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//BiasedMF"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File("conf//biasedMF.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); logger.flush(); DataLoaderFile loader = new DataLoaderFile(".//data//MoveLens100k.txt"); loader.readSimple(); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; double totalPrecision = 0; double totalRecall = 0; double totalMAP = 0; double totalNDCG = 0; double totalMRR = 0; double totalAUC = 0; int F = 5; logger.println(F + "- folder cross validation."); logger.flush(); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { System.out.println("Folder: " + folder); logger.println("Folder: " + folder); logger.flush(); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = new HashMap<String, Integer>(); HashMap<String, Integer> itemIDIndexMapping = new HashMap<String, Integer>(); for (int i = 0; i < dataset.getUserIDs().size(); i++) { userIDIndexMapping.put(dataset.getUserIDs().get(i), i); } for (int i = 0; i < dataset.getItemIDs().size(); i++) { itemIDIndexMapping.put(dataset.getItemIDs().get(i), i); } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { // if( testRatings.get(i).getValue() < 5 ) // continue; testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } System.out.println("Training: " + trainRatingMatrix.getTotalRatingNumber() + " vs Test: " + testRatingMatrix.getTotalRatingNumber()); logger.println("Initialize a biased matrix factorization recommendation model."); logger.flush(); BiasedMF algo = new BiasedMF(trainRatingMatrix, false, ".//localModels//" + config.getString("NAME")); algo.setLogger(logger); algo.build(); algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); logger.flush(); //rating prediction accuracy double RMSE = 0; double MAE = 0; double precision = 0; double recall = 0; double map = 0; double ndcg = 0; double mrr = 0; double auc = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID()), false); if (prediction > algo.getMaxRating()) prediction = algo.getMaxRating(); if (prediction < algo.getMinRating()) prediction = algo.getMinRating(); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; System.out.println("Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Folder --- MAE: " + MAE + " RMSE: " + RMSE); //ranking accuracy if (algo.getTopN() > 0) { HashMap<Integer, ArrayList<ResultUnit>> results = new HashMap<Integer, ArrayList<ResultUnit>>(); for (int i = 0; i < trainRatingMatrix.getRow(); i++) { ArrayList<ResultUnit> rec = algo.getRecommendationList(i); if (rec == null) continue; int total = testRatingMatrix.getUserRatingNumber(i); if (total == 0)//this user is ignored continue; results.put(i, rec); } RankResultGenerator generator = new RankResultGenerator(results, algo.getTopN(), testRatingMatrix, trainRatingMatrix); precision = generator.getPrecisionN(); totalPrecision = totalPrecision + precision; recall = generator.getRecallN(); totalRecall = totalRecall + recall; map = generator.getMAPN(); totalMAP = totalMAP + map; ndcg = generator.getNDCGN(); totalNDCG = totalNDCG + ndcg; mrr = generator.getMRRN(); totalMRR = totalMRR + mrr; auc = generator.getAUC(); totalAUC = totalAUC + auc; System.out.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); logger.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); } logger.flush(); } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); System.out.println("Precision@N: " + totalPrecision / F); System.out.println("Recall@N: " + totalRecall / F); System.out.println("MAP@N: " + totalMAP / F); System.out.println("MRR@N: " + totalMRR / F); System.out.println("NDCG@N: " + totalNDCG / F); System.out.println("AUC@N: " + totalAUC / F); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + "\n" + "MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F + "\n" + "Precision@N: " + totalPrecision / F + "\n" + "Recall@N: " + totalRecall / F + "\n" + "MAP@N: " + totalMAP / F + "\n" + "MRR@N: " + totalMRR / F + "\n" + "NDCG@N: " + totalNDCG / F + "\n" + "AUC@N: " + totalAUC / F); logger.flush(); logger.close(); }
From source file:ch.epfl.lsir.xin.test.MFTest.java
/** * @param args//from w w w . j a va 2 s.c o m */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//MF"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File("conf//MF.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); logger.flush(); DataLoaderFile loader = new DataLoaderFile(".//data//MoveLens100k.txt"); loader.readSimple(); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; double totalPrecision = 0; double totalRecall = 0; double totalMAP = 0; double totalNDCG = 0; double totalMRR = 0; double totalAUC = 0; int F = 5; logger.println(F + "- folder cross validation."); logger.flush(); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { System.out.println("Folder: " + folder); logger.println("Folder: " + folder); logger.flush(); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = new HashMap<String, Integer>(); HashMap<String, Integer> itemIDIndexMapping = new HashMap<String, Integer>(); for (int i = 0; i < dataset.getUserIDs().size(); i++) { userIDIndexMapping.put(dataset.getUserIDs().get(i), i); } for (int i = 0; i < dataset.getItemIDs().size(); i++) { itemIDIndexMapping.put(dataset.getItemIDs().get(i), i); } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { // if( testRatings.get(i).getValue() < 5 ) // continue; testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } System.out.println("Training: " + trainRatingMatrix.getTotalRatingNumber() + " vs Test: " + testRatingMatrix.getTotalRatingNumber()); logger.println("Initialize a matrix factorization based recommendation model."); logger.flush(); MatrixFactorization algo = new MatrixFactorization(trainRatingMatrix, false, ".//localModels//" + config.getString("NAME")); algo.setLogger(logger); algo.build(); algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); logger.flush(); //rating prediction accuracy double RMSE = 0; double MAE = 0; double precision = 0; double recall = 0; double map = 0; double ndcg = 0; double mrr = 0; double auc = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID()), false); if (prediction > algo.getMaxRating()) prediction = algo.getMaxRating(); if (prediction < algo.getMinRating()) prediction = algo.getMinRating(); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; System.out.println("Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Folder --- MAE: " + MAE + " RMSE: " + RMSE); //ranking accuracy if (algo.getTopN() > 0) { HashMap<Integer, ArrayList<ResultUnit>> results = new HashMap<Integer, ArrayList<ResultUnit>>(); for (int i = 0; i < trainRatingMatrix.getRow(); i++) { ArrayList<ResultUnit> rec = algo.getRecommendationList(i); if (rec == null) continue; int total = testRatingMatrix.getUserRatingNumber(i); if (total == 0)//this user is ignored continue; results.put(i, rec); // for( Map.Entry<Integer, Double> entry : testRatingMatrix.getRatingMatrix().get(i).entrySet() ) // { // System.out.print( entry.getKey() + "(" + entry.getValue() + ") , "); // } // System.out.println(); // for( int j = 0 ; j < rec.size() ; j++ ) // { // System.out.print(rec.get(j).getItemIndex() + "(" + rec.get(j).getPrediciton() + // ") , "); // } // System.out.println("**********"); } RankResultGenerator generator = new RankResultGenerator(results, algo.getTopN(), testRatingMatrix, trainRatingMatrix); precision = generator.getPrecisionN(); totalPrecision = totalPrecision + precision; recall = generator.getRecallN(); totalRecall = totalRecall + recall; map = generator.getMAPN(); totalMAP = totalMAP + map; ndcg = generator.getNDCGN(); totalNDCG = totalNDCG + ndcg; mrr = generator.getMRRN(); totalMRR = totalMRR + mrr; auc = generator.getAUC(); totalAUC = totalAUC + auc; System.out.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); logger.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " MAE: " + MAE + " RMSE: " + RMSE); logger.flush(); } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); System.out.println("Precision@N: " + totalPrecision / F); System.out.println("Recall@N: " + totalRecall / F); System.out.println("MAP@N: " + totalMAP / F); System.out.println("MRR@N: " + totalMRR / F); System.out.println("NDCG@N: " + totalNDCG / F); System.out.println("AUC@N: " + totalAUC / F); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + "\n" + "MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F + "\n" + "Precision@N: " + totalPrecision / F + "\n" + "Recall@N: " + totalRecall / F + "\n" + "MAP@N: " + totalMAP / F + "\n" + "MRR@N: " + totalMRR / F + "\n" + "NDCG@N: " + totalNDCG / F + "\n" + "AUC@N: " + totalAUC / F); logger.flush(); logger.close(); }
From source file:ch.epfl.lsir.xin.test.UserBasedCFTest.java
/** * @param args//w ww. j a v a 2 s . c o m */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//UserBasedCF"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File(".//conf//UserBasedCF.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); DataLoaderFile loader = new DataLoaderFile(".//data//MoveLens100k.txt"); loader.readSimple(); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; double totalPrecision = 0; double totalRecall = 0; double totalMAP = 0; double totalNDCG = 0; double totalMRR = 0; double totalAUC = 0; int F = 5; logger.println(F + "- folder cross validation."); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { logger.println("Folder: " + folder); System.out.println("Folder: " + folder); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = new HashMap<String, Integer>(); HashMap<String, Integer> itemIDIndexMapping = new HashMap<String, Integer>(); for (int i = 0; i < dataset.getUserIDs().size(); i++) { userIDIndexMapping.put(dataset.getUserIDs().get(i), i); } for (int i = 0; i < dataset.getItemIDs().size(); i++) { itemIDIndexMapping.put(dataset.getItemIDs().get(i), i); } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } trainRatingMatrix.calculateGlobalAverage(); trainRatingMatrix.calculateUsersMean(); RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { // if( testRatings.get(i).getValue() < 5 ) // continue; testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } logger.println("Initialize a user based collaborative filtering recommendation model."); UserBasedCF algo = new UserBasedCF(trainRatingMatrix, false, ".//localModels//" + config.getString("NAME")); algo.setLogger(logger); algo.build();//if read local model, no need to build the model algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); System.out.println(trainRatings.size() + " vs. " + testRatings.size()); logger.flush(); //rating prediction accuracy double RMSE = 0; double MAE = 0; double precision = 0; double recall = 0; double map = 0; double ndcg = 0; double mrr = 0; double auc = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID()), false); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } if (prediction > algo.getMaxRating()) prediction = algo.getMaxRating(); if (prediction < algo.getMinRating()) prediction = algo.getMinRating(); MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; System.out.println("Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.flush(); //ranking accuracy if (algo.getTopN() > 0) { HashMap<Integer, ArrayList<ResultUnit>> results = new HashMap<Integer, ArrayList<ResultUnit>>(); for (int i = 0; i < testRatingMatrix.getRow(); i++) { ArrayList<ResultUnit> rec = algo.getRecommendationList(i); if (rec == null) continue; int total = testRatingMatrix.getUserRatingNumber(i); if (total == 0)//this user is ignored continue; results.put(i, rec); // for( Map.Entry<Integer, Double> entry : testRatingMatrix.getRatingMatrix().get(i).entrySet() ) // { // System.out.print( entry.getKey() + "(" + entry.getValue() + ") , "); // } // System.out.println(); // for( int j = 0 ; j < rec.size() ; j++ ) // { // System.out.print(rec.get(j).getItemIndex() + "(" + rec.get(j).getPrediciton() + // ") , "); // } // System.out.println("**********"); } RankResultGenerator generator = new RankResultGenerator(results, algo.getTopN(), testRatingMatrix, trainRatingMatrix); precision = generator.getPrecisionN(); totalPrecision = totalPrecision + precision; recall = generator.getRecallN(); totalRecall = totalRecall + recall; map = generator.getMAPN(); totalMAP = totalMAP + map; ndcg = generator.getNDCGN(); totalNDCG = totalNDCG + ndcg; mrr = generator.getMRRN(); totalMRR = totalMRR + mrr; auc = generator.getAUC(); totalAUC = totalAUC + auc; System.out.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); logger.println("Folder --- precision: " + precision + " recall: " + recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); } } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); System.out.println("Precision@N: " + totalPrecision / F); System.out.println("Recall@N: " + totalRecall / F); System.out.println("MAP@N: " + totalMAP / F); System.out.println("MRR@N: " + totalMRR / F); System.out.println("NDCG@N: " + totalNDCG / F); System.out.println("AUC@N: " + totalAUC / F); // MovieLens100k //MAE: 0.7343907480119425 RMSE: 0.9405808357192891 (MovieLens 100K, shrinkage 25, neighbor size 60, PCC) //MAE: 0.7522376630596646 RMSE: 0.9520931265724659 (MovieLens 100K, no shrinkage , neighbor size 40, COSINE) logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + "\n" + "MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F + "\n" + "Precision@N: " + totalPrecision / F + "\n" + "Recall@N: " + totalRecall / F + "\n" + "MAP@N: " + totalMAP / F + "\n" + "MRR@N: " + totalMRR / F + "\n" + "NDCG@N: " + totalNDCG / F + "\n" + "AUC@N: " + totalAUC / F); logger.flush(); logger.close(); }
From source file:edu.upenn.egricelab.AlignerBoost.FilterSAMAlignPE.java
public static void main(String[] args) { if (args.length == 0) { printUsage();/*w w w . ja v a 2s. co m*/ return; } try { parseOptions(args); } catch (IllegalArgumentException e) { System.err.println("Error: " + e.getMessage()); printUsage(); return; } // Read in chrList, if specified if (chrFile != null) { chrFilter = new HashSet<String>(); try { BufferedReader chrFilterIn = new BufferedReader(new FileReader(chrFile)); String chr = null; while ((chr = chrFilterIn.readLine()) != null) chrFilter.add(chr); chrFilterIn.close(); if (verbose > 0) System.err.println( "Only looking at alignments on " + chrFilter.size() + " specified chromosomes"); } catch (IOException e) { System.err.println("Error: " + e.getMessage()); return; } } if (verbose > 0) { // Start the processMonitor processMonitor = new Timer(); // Start the ProcessStatusTask statusTask = new ProcessStatusTask(); // Schedule to show the status every 1 second processMonitor.scheduleAtFixedRate(statusTask, 0, statusFreq); } // Read in known SNP file, if specified if (knownSnpFile != null) { if (verbose > 0) System.err.println("Checking known SNPs from user specified VCF file"); knownVCF = new VCFFileReader(new File(knownSnpFile)); } SamReaderFactory readerFac = SamReaderFactory.makeDefault(); SAMFileWriterFactory writerFac = new SAMFileWriterFactory(); if (!isSilent) readerFac.validationStringency(ValidationStringency.LENIENT); // use LENIENT stringency else readerFac.validationStringency(ValidationStringency.SILENT); // use SILENT stringency SamReader in = readerFac.open(new File(inFile)); SAMFileHeader inHeader = in.getFileHeader(); if (inHeader.getGroupOrder() == GroupOrder.reference && inHeader.getSortOrder() == SortOrder.coordinate) System.err.println("Warning: Input file '" + inFile + "' might be sorted by coordinate and cannot be correctly processed!"); SAMFileHeader header = inHeader.clone(); // copy the inFile header as outFile header // Add new programHeader SAMProgramRecord progRec = new SAMProgramRecord(progName); progRec.setProgramName(progName); progRec.setProgramVersion(progVer); progRec.setCommandLine(StringUtils.join(" ", args)); header.addProgramRecord(progRec); //System.err.println(inFile + " groupOrder: " + in.getFileHeader().getGroupOrder() + " sortOrder: " + in.getFileHeader().getSortOrder()); // reset the orders header.setGroupOrder(groupOrder); header.setSortOrder(sortOrder); // write SAMHeader String prevID = null; SAMRecord prevRecord = null; List<SAMRecord> alnList = new ArrayList<SAMRecord>(); List<SAMRecordPair> alnPEList = null; // Estimate fragment length distribution by scan one-pass through the alignments SAMRecordIterator results = in.iterator(); if (!NO_ESTIMATE) { if (verbose > 0) { System.err.println("Estimating insert fragment size distribution ..."); statusTask.reset(); statusTask.setInfo("alignments scanned"); } long N = 0; double fragL_S = 0; // fragLen sum double fragL_SS = 0; // fragLen^2 sum while (results.hasNext()) { SAMRecord record = results.next(); if (verbose > 0) statusTask.updateStatus(); if (record.getFirstOfPairFlag() && !record.isSecondaryOrSupplementary()) { double fragLen = Math.abs(record.getInferredInsertSize()); if (fragLen != 0 && fragLen >= MIN_FRAG_LEN && fragLen <= MAX_FRAG_LEN) { // only consider certain alignments N++; fragL_S += fragLen; fragL_SS += fragLen * fragLen; } // stop estimate if already enough if (MAX_ESTIMATE_SCAN > 0 && N >= MAX_ESTIMATE_SCAN) break; } } if (verbose > 0) statusTask.finish(); // estimate fragment size if (N >= MIN_ESTIMATE_BASE) { // override command line values MEAN_FRAG_LEN = fragL_S / N; SD_FRAG_LEN = Math.sqrt((N * fragL_SS - fragL_S * fragL_S) / (N * (N - 1))); String estStr = String.format("Estimated fragment size distribution: N(%.1f, %.1f)", MEAN_FRAG_LEN, SD_FRAG_LEN); if (verbose > 0) System.err.println(estStr); // also add the estimation to comment header.addComment(estStr); } else { System.err.println( "Unable to estimate the fragment size distribution due to too few observed alignments"); System.err.println( "You have to specify the '--mean-frag-len' and '--sd-frag-len' on the command line and re-run this step"); statusTask.cancel(); processMonitor.cancel(); return; } // Initiate the normal model normModel = new NormalDistribution(MEAN_FRAG_LEN, SD_FRAG_LEN); // reset the iterator, if necessary if (in.type() == SamReader.Type.SAM_TYPE) { try { in.close(); } catch (IOException e) { System.err.println(e.getMessage()); } in = readerFac.open(new File(inFile)); } results.close(); results = in.iterator(); } // end of NO_ESTIMATE SAMFileWriter out = OUT_IS_SAM ? writerFac.makeSAMWriter(header, false, new File(outFile)) : writerFac.makeBAMWriter(header, false, new File(outFile)); // check each alignment again if (verbose > 0) { System.err.println("Filtering alignments ..."); statusTask.reset(); statusTask.setInfo("alignments processed"); } while (results.hasNext()) { SAMRecord record = results.next(); if (verbose > 0) statusTask.updateStatus(); String ID = record.getReadName(); // fix read and quality string for this read, if is a secondary hit from multiple hits, used for BWA alignment if (ID.equals(prevID) && record.getReadLength() == 0) SAMAlignFixer.fixSAMRecordRead(record, prevRecord); if (chrFilter != null && !chrFilter.contains(record.getReferenceName())) { prevID = ID; prevRecord = record; continue; } // fix MD:Z string for certain aligners with invalid format (i.e. seqAlto) if (fixMD) SAMAlignFixer.fixMisStr(record); // fix alignment, ignore if failed (unmapped or empty) if (!SAMAlignFixer.fixSAMRecord(record, knownVCF, DO_1DP)) { prevID = ID; prevRecord = record; continue; } if (!record.getReadPairedFlag()) { System.err.println("Error: alignment is not from a paired-end read at\n" + record.getSAMString()); out.close(); statusTask.cancel(); processMonitor.cancel(); return; } if (!ID.equals(prevID) && prevID != null || !results.hasNext()) { // a non-first new ID meet, or end of alignments // create alnPEList from filtered alnList alnPEList = createAlnPEListFromAlnList(alnList); //System.err.printf("%d alignments for %s transformed to %d alnPairs%n", alnList.size(), prevID, alnPEList.size()); int totalPair = alnPEList.size(); // filter highly unlikely PEhits filterPEHits(alnPEList, MIN_ALIGN_RATE, MIN_IDENTITY); // calculate posterior mapQ for each pair calcPEHitPostP(alnPEList, totalPair, MAX_HIT); // filter hits by mapQ if (MIN_MAPQ > 0) filterPEHits(alnPEList, MIN_MAPQ); // sort the list first with an anonymous class of comparator, with DESCREASING order Collections.sort(alnPEList, Collections.reverseOrder()); // control max-best if (MAX_BEST != 0 && alnPEList.size() > MAX_BEST) { // potential too much best hits int nBestStratum = 0; int bestMapQ = alnPEList.get(0).getPEMapQ(); // best mapQ from first PE for (SAMRecordPair pr : alnPEList) if (pr.getPEMapQ() == bestMapQ) nBestStratum++; else break; // stop searching for sorted list if (nBestStratum > MAX_BEST) alnPEList.clear(); } // filter alignments with auxiliary filters if (!MAX_SENSITIVITY) filterPEHits(alnPEList, MAX_SEED_MIS, MAX_SEED_INDEL, MAX_ALL_MIS, MAX_ALL_INDEL); // report remaining secondary alignments, up-to MAX_REPORT for (int i = 0; i < alnPEList.size() && (MAX_REPORT == 0 || i < MAX_REPORT); i++) { SAMRecordPair repPair = alnPEList.get(i); if (doUpdateBit) repPair.setNotPrimaryAlignmentFlags(i != 0); int nReport = MAX_REPORT == 0 ? Math.min(alnPEList.size(), MAX_REPORT) : alnPEList.size(); int nFiltered = alnPEList.size(); if (repPair.fwdRecord != null) { repPair.fwdRecord.setAttribute("NH", nReport); repPair.fwdRecord.setAttribute("XN", nFiltered); out.addAlignment(repPair.fwdRecord); } if (repPair.revRecord != null) { repPair.revRecord.setAttribute("NH", nReport); repPair.revRecord.setAttribute("XN", nFiltered); out.addAlignment(repPair.revRecord); } } // reset list alnList.clear(); alnPEList.clear(); } // update if (!ID.equals(prevID)) { prevID = ID; prevRecord = record; } alnList.add(record); } // end while try { in.close(); out.close(); } catch (IOException e) { System.err.println(e.getMessage()); } // Terminate the monitor task and monitor if (verbose > 0) { statusTask.cancel(); statusTask.finish(); processMonitor.cancel(); } }
From source file:ch.epfl.lsir.xin.test.SocialRegTest.java
/** * @param args//w w w . j ava 2s .c o m */ public static void main(String[] args) throws Exception { // TODO Auto-generated method stub PrintWriter logger = new PrintWriter(".//results//SocialReg"); PropertiesConfiguration config = new PropertiesConfiguration(); config.setFile(new File("conf//SocialReg.properties")); try { config.load(); } catch (ConfigurationException e) { // TODO Auto-generated catch block e.printStackTrace(); } logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Read rating data..."); logger.flush(); DataLoaderFile loader = new DataLoaderFile(".//data//Epinions-ratings.txt"); loader.readSimple(); //read social information loader.readRelation(".//data//Epinions-trust.txt"); DataSetNumeric dataset = loader.getDataset(); System.out.println("Number of ratings: " + dataset.getRatings().size() + " Number of users: " + dataset.getUserIDs().size() + " Number of items: " + dataset.getItemIDs().size()); logger.println("Number of ratings: " + dataset.getRatings().size() + ", Number of users: " + dataset.getUserIDs().size() + ", Number of items: " + dataset.getItemIDs().size()); logger.flush(); double totalMAE = 0; double totalRMSE = 0; double totalPrecision = 0; double totalRecall = 0; double totalMAP = 0; double totalNDCG = 0; double totalMRR = 0; double totalAUC = 0; int F = 5; logger.println(F + "- folder cross validation."); logger.flush(); ArrayList<ArrayList<NumericRating>> folders = new ArrayList<ArrayList<NumericRating>>(); for (int i = 0; i < F; i++) { folders.add(new ArrayList<NumericRating>()); } while (dataset.getRatings().size() > 0) { int index = new Random().nextInt(dataset.getRatings().size()); int r = new Random().nextInt(F); folders.get(r).add(dataset.getRatings().get(index)); dataset.getRatings().remove(index); } for (int folder = 1; folder <= F; folder++) { System.out.println("Folder: " + folder); logger.println("Folder: " + folder); logger.flush(); ArrayList<NumericRating> trainRatings = new ArrayList<NumericRating>(); ArrayList<NumericRating> testRatings = new ArrayList<NumericRating>(); for (int i = 0; i < folders.size(); i++) { if (i == folder - 1)//test data { testRatings.addAll(folders.get(i)); } else {//training data trainRatings.addAll(folders.get(i)); } } //create rating matrix HashMap<String, Integer> userIDIndexMapping = dataset.getUserIDMapping(); HashMap<String, Integer> itemIDIndexMapping = dataset.getItemIDMapping(); // for( int i = 0 ; i < dataset.getUserIDs().size() ; i++ ) // { // userIDIndexMapping.put(dataset.getUserIDs().get(i), i); // } // for( int i = 0 ; i < dataset.getItemIDs().size() ; i++ ) // { // itemIDIndexMapping.put(dataset.getItemIDs().get(i) , i); // } RatingMatrix trainRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < trainRatings.size(); i++) { trainRatingMatrix.set(userIDIndexMapping.get(trainRatings.get(i).getUserID()), itemIDIndexMapping.get(trainRatings.get(i).getItemID()), trainRatings.get(i).getValue()); } RatingMatrix testRatingMatrix = new RatingMatrix(dataset.getUserIDs().size(), dataset.getItemIDs().size()); for (int i = 0; i < testRatings.size(); i++) { testRatingMatrix.set(userIDIndexMapping.get(testRatings.get(i).getUserID()), itemIDIndexMapping.get(testRatings.get(i).getItemID()), testRatings.get(i).getValue()); } System.out.println("Training: " + trainRatingMatrix.getTotalRatingNumber() + " vs Test: " + testRatingMatrix.getTotalRatingNumber()); logger.println("Initialize a social regularization recommendation model."); logger.flush(); SocialReg algo = new SocialReg(trainRatingMatrix, dataset.getRelationships(), false, ".//localModels//" + config.getString("NAME")); algo.setLogger(logger); algo.build(); algo.saveModel(".//localModels//" + config.getString("NAME")); logger.println("Save the model."); logger.flush(); System.out.println(trainRatings.size() + " vs. " + testRatings.size()); //rating prediction accuracy double RMSE = 0; double MAE = 0; double precision = 0; double recall = 0; double map = 0; double ndcg = 0; double mrr = 0; double auc = 0; int count = 0; for (int i = 0; i < testRatings.size(); i++) { NumericRating rating = testRatings.get(i); double prediction = algo.predict(userIDIndexMapping.get(rating.getUserID()), itemIDIndexMapping.get(rating.getItemID())); if (prediction > algo.getMaxRating()) prediction = algo.getMaxRating(); if (prediction < algo.getMinRating()) prediction = algo.getMinRating(); if (Double.isNaN(prediction)) { System.out.println("no prediction"); continue; } MAE = MAE + Math.abs(rating.getValue() - prediction); RMSE = RMSE + Math.pow((rating.getValue() - prediction), 2); count++; } MAE = MAE / count; RMSE = Math.sqrt(RMSE / count); totalMAE = totalMAE + MAE; totalRMSE = totalRMSE + RMSE; System.out.println("Folder --- MAE: " + MAE + " RMSE: " + RMSE); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + " Folder --- MAE: " + MAE + " RMSE: " + RMSE); //ranking accuracy // if( algo.getTopN() > 0 ) // { // HashMap<Integer , ArrayList<ResultUnit>> results = new HashMap<Integer , ArrayList<ResultUnit>>(); // for( int i = 0 ; i < trainRatingMatrix.getRow() ; i++ ) // { // ArrayList<ResultUnit> rec = algo.getRecommendationList(i); // results.put(i, rec); // } // RankResultGenerator generator = new RankResultGenerator(results , algo.getTopN() , testRatingMatrix); // precision = generator.getPrecisionN(); // totalPrecision = totalPrecision + precision; // recall = generator.getRecallN(); // totalRecall = totalRecall + recall; // map = generator.getMAPN(); // totalMAP = totalMAP + map; // ndcg = generator.getNDCGN(); // totalNDCG = totalNDCG + ndcg; // mrr = generator.getMRRN(); // totalMRR = totalMRR + mrr; // auc = generator.getAUC(); // totalAUC = totalAUC + auc; // System.out.println("Folder --- precision: " + precision + " recall: " + // recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + mrr + " auc: " + auc); // logger.println("Folder --- precision: " + precision + " recall: " + // recall + " map: " + map + " ndcg: " + ndcg + " mrr: " + // mrr + " auc: " + auc); // } logger.flush(); } System.out.println("MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F); System.out.println("Precision@N: " + totalPrecision / F); System.out.println("Recall@N: " + totalRecall / F); System.out.println("MAP@N: " + totalMAP / F); System.out.println("MRR@N: " + totalMRR / F); System.out.println("NDCG@N: " + totalNDCG / F); System.out.println("AUC@N: " + totalAUC / F); logger.println(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date()) + "\n" + "MAE: " + totalMAE / F + " RMSE: " + totalRMSE / F + "\n" + "Precision@N: " + totalPrecision / F + "\n" + "Recall@N: " + totalRecall / F + "\n" + "MAP@N: " + totalMAP / F + "\n" + "MRR@N: " + totalMRR / F + "\n" + "NDCG@N: " + totalNDCG / F + "\n" + "AUC@N: " + totalAUC / F); logger.flush(); logger.close(); }
From source file:eu.crisis_economics.abm.model.configuration.LogNormalDistributionModelParameterConfiguration.java
/** * A lightweight test for this {@link ConfigurationComponent}. This snippet * creates a {@link Parameter}{@code <Double>} using an instance of * {@link LogNormalDistributionModelParameterConfiguration} and then tests * whether the logarithm of {@link Double} values drawn from this {@link Parameter} * have the expected mean.//from w w w .jav a 2s. c om */ public static void main(String[] args) { { final LogNormalDistributionModelParameterConfiguration configuration = new LogNormalDistributionModelParameterConfiguration(); configuration.setMu(5.0); configuration.setSigma(0.1); final ModelParameter<Double> distribution = configuration.createInjector() .getInstance(Key.get(new TypeLiteral<ModelParameter<Double>>() { })); double mean = 0.; final int numSamples = 10000000; for (int i = 0; i < numSamples; ++i) { final double value = Math.log(distribution.get()); mean += value; } mean /= numSamples; Assert.assertTrue(Math.abs(mean - 5.) < 1.e-3); } { final LogNormalDistributionModelParameterConfiguration configuration = LogNormalDistributionModelParameterConfiguration .create(5., .1); final ModelParameter<Double> distribution = configuration.createInjector() .getInstance(Key.get(new TypeLiteral<ModelParameter<Double>>() { })); final Variance variance = new Variance(); double mean = 0.; final int numSamples = 10000000; for (int i = 0; i < numSamples; ++i) { final double value = distribution.get(); mean += value; variance.increment(value); } mean /= numSamples; final double observedSigma = Math.sqrt(variance.getResult()); Assert.assertTrue(Math.abs(mean - 5.) < 1.e-3); Assert.assertTrue(Math.abs(observedSigma - .1) < 1.e-3); } }