logistic-regression-movie-review-classifier.ipynb

{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load in \n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nfrom scipy.sparse import csr_matrix\nfrom tqdm import tqdm, trange\nfrom sklearn.datasets import load_svmlight_file\n# Input data files are available in the \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n\nimport os\nprint(os.listdir(\"../input/movie-ratings\"))\n\n# Any results you write to the current directory are saved as output.","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true},"cell_type":"code","source":"training_data, training_labels = load_svmlight_file('../input/movie-ratings/movie-ratings/data-splits/data.train')\ntesting_data, testing_labels = load_svmlight_file('../input/movie-ratings/movie-ratings/data-splits/data.test')\ntesting_data = csr_matrix((testing_data.data, testing_data.indices, testing_data.indptr), shape=(testing_data.shape[0], training_data.shape[1]))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"1251c31a55706de89be3889b920f40f75ec6f20e"},"cell_type":"code","source":"print(training_data.shape, training_labels.shape)\nprint(testing_data.shape, testing_labels.shape)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"3609f58bf349c7b87265e8427a29e8029836140b"},"cell_type":"code","source":"class Softmax:\n    def __init__(self, learning_rate=0.1, reg_const=0.05, num_iters=10, batch_size=200):\n        self.W = None\n        self.learning_rate =  learning_rate\n        self.reg_const =  reg_const\n        self.num_iters = num_iters\n        self.batch_size = batch_size\n        \n    def loss(self,X_batch, y_batch):\n        \n        num_train = X_batch.shape[0]\n        num_classes = self.W.shape[1]\n        \n        scores = X_batch.dot(self.W)\n        shift_scores = scores - np.max(scores, axis = 1).reshape(-1,1)\n        softmax_output = np.exp(shift_scores)/np.sum(np.exp(shift_scores), axis = 1).reshape(-1,1)\n        loss = -np.sum(np.log(softmax_output[range(num_train), y_batch]))\n        loss += 0.5 * self.reg_const * np.sum(self.W*self.W)\n        \n        dS = softmax_output.copy()\n        dS[range(num_train), y_batch] += -1\n        dW = (X_batch.T).dot(dS)\n        dW = dW/num_train + self.reg_const * self.W\n        \n        return loss, dW\n    \n    def fit(self, X, y, **kwargs):\n        if kwargs:\n            self.learning_rate =  kwargs['learning_rate']\n            self.reg_const =  kwargs['reg_const']\n            self.num_iters = kwargs['num_iters']\n            self.batch_size = kwargs['batch_size']\n        \n        num_train, dim = X.shape\n        num_classes = 2 # assume y takes values 0...K-1 where K is number of classes\n        if self.W is None:\n          # lazily initialize W\n          self.W = 0.001 * np.random.randn(int(dim), int(num_classes))\n\n        # Run stochastic gradient descent to optimize W\n        for it in trange(self.num_iters, mininterval=5):\n            batch_idx = np.random.choice(num_train, self.batch_size, replace = True)\n            X_batch =  X[batch_idx]\n            y_batch = y[batch_idx]\n        \n            loss, grad = self.loss(X_batch, y_batch.astype(int))\n        \n            self.W += - self.learning_rate * grad\n    \n    def get_params(self, deep = False):\n        return {'learning_rate': self.learning_rate, 'reg_const': self.reg_const, \n                'num_iters':self.num_iters, 'batch_size':self.batch_size}\n    \n    def set_params(self, **parameters):\n        for parameter, value in parameters.items():\n            setattr(self, parameter, value)\n        return self\n    \n    def predict(self, X):\n        scores = X.dot(self.W)\n        y_pred = np.argmax(scores, axis =1)\n        return y_pred\n    \n    def score(self, X, y):\n        assert X.shape[0] == y.shape[0]\n        correct = sum(1 for actual, prediction in zip(y, self.predict(X)) if np.sign(actual) == np.sign(prediction))\n        return correct/X.shape[0]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"1175051653fb97909c327e8022ebfd31e9d9c671","_kg_hide-output":true,"scrolled":false},"cell_type":"code","source":"softmax = Softmax()\nsoftmax.fit(training_data, training_labels, learning_rate=0.01, reg_const=0.05, num_iters=20000, batch_size=200)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"dbebcb08c80a08fb81f2e4df6a4bdc4b57df3460"},"cell_type":"code","source":"softmax.score(testing_data, testing_labels)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"4dd2ad73bd5a915a643361653bac2cb865632555"},"cell_type":"code","source":"from sklearn.model_selection import cross_val_score, GridSearchCV\nclf = GridSearchCV(Softmax(), param_grid={'learning_rate':[0.0001, 0.0005, 0.001], 'reg_const':[0.0001,0.001,0.01,0.1], 'num_iters':[20000], 'batch_size':[200]}, cv=5, scoring='accuracy', n_jobs=-1)\nclf.fit(testing_data, testing_labels)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"db3835e865350967d29d74c9486cc5abc20d9214"},"cell_type":"code","source":"print(clf.best_params_)\nprint(clf.best_score_)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"d3ce5941c7faaeaf58ade54f7f946f48550f2fab"},"cell_type":"code","source":"cross_val_score(Softmax(), training_data, y=training_labels, scoring='accuracy', \n                fit_params=clf.best_params_, cv=5, n_jobs=-1)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"670171e3d4904143b7a73de17a8be4e907d2f1f9"},"cell_type":"code","source":"clf.best_estimator_.score(testing_data, testing_labels)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"9f62f00f186c374e0735c409978b152ee06c4e80"},"cell_type":"code","source":"softmax = clf.best_estimator_\neval_data, _ = load_svmlight_file('../input/movie-ratings/movie-ratings/data-splits/data.eval.anon')\neval_data = csr_matrix((eval_data.data, eval_data.indices, eval_data.indptr), shape=(eval_data.shape[0], training_data.shape[1]))\nsubmission_pred = softmax.predict(eval_data)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"8e72b1320d005ee1fe8a60e84b2de58b7abca859","scrolled":false},"cell_type":"code","source":"with open('submission.csv', 'w') as submission:\n    with open('../input/movie-ratings/movie-ratings/data-splits/data.eval.anon.id', 'r') as example_ids:\n        submission.write('example_id,label\\n')\n        for example_id, label in zip(example_ids, submission_pred):\n            submission.write('{},{}\\n'.format(example_id.strip(), int(label)))","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}