{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"import numpy as np\nimport pandas as pd\nimport os\n\nimport matplotlib.pyplot as plt\n%matplotlib inline\n","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Use first 10**5 rows of train dataset for data exploring. Using more efficient datatypes as shown in introduction notebook."},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"train_df_105 = pd.read_csv('/kaggle/input/riiid-test-answer-prediction/train.csv', low_memory=False, nrows=10**5, \n                       dtype={'row_id': 'int64', 'timestamp': 'int64', 'user_id': 'int32', 'content_id': 'int16', 'content_type_id': 'int8',\n                              'task_container_id': 'int16', 'user_answer': 'int8', 'answered_correctly': 'int8', 'prior_question_elapsed_time': 'float32', \n                             'prior_question_had_explanation': 'boolean',\n                             }\n                      )","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#data ordered by uers and timestamp, so we drop last user\ntrain_df_105.drop(train_df_105[train_df_105.user_id == train_df_105.user_id.iloc[-1]].index, inplace=True)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Exploring questions stats"},{"metadata":{"trusted":true},"cell_type":"code","source":"#only questions without lectures\ntrain_df_105_quest = train_df_105[train_df_105.content_type_id ==0]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#correct answers percentage by user\nuser_quest_stats = train_df_105_quest.groupby('user_id')['answered_correctly'].agg(correct_answers_percentage='mean')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"user_quest_stats.plot.hist(bins=100)\nplt.title(\"Correct ansewrs percentage distribution by users\")\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest[train_df_105_quest.answered_correctly == 1].prior_question_elapsed_time.plot.hist(bins=100, label='correct')\ntrain_df_105_quest[train_df_105_quest.answered_correctly == 0].prior_question_elapsed_time.plot.hist(bins=100, label='incorrect')\nplt.title(\"Correct and incorrect elapsed time\")\nplt.legend()\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest.groupby('prior_question_had_explanation')['answered_correctly'].agg(correct_answers_percentage='mean')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest['answered_correctly'].agg(correct_answers_percentage='mean')","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Exploring questions and lectures"},{"metadata":{"trusted":true},"cell_type":"code","source":"questions_df = pd.read_csv('/kaggle/input/riiid-test-answer-prediction/questions.csv')\nlectures_df = pd.read_csv('/kaggle/input/riiid-test-answer-prediction/lectures.csv')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"questions_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest_with_questions = train_df_105_quest.merge(questions_df,left_on='content_id', right_on='question_id')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest_with_questions['tags_list'] = [x.split() for x in train_df_105_quest_with_questions.tags.values]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df_105_quest_with_questions.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"part_correct_answers=train_df_105_quest_with_questions.groupby('part')['answered_correctly'].agg(correct_answers_percentage='mean')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"part_correct_answers.plot.line()\nplt.title(\"Correct ansewrs in parts\")\nplt.ylim(0,1)\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"tags_correct_answers = train_df_105_quest_with_questions[['answered_correctly', 'tags_list']]\ntags_correct_answers = tags_correct_answers.explode('tags_list')\ntags_correct_answers = tags_correct_answers.rename(columns={'tags_list':'tag'})","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"tags_correct_answers_percentage=tags_correct_answers.groupby('tag')['answered_correctly'].agg(correct_answers_percentage='mean')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"tags_correct_answers_percentage.hist(bins=100)\nplt.title(\"Correct ansewrs percentage distribution by tags\")\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"import lightgbm as lgb\nfrom sklearn.metrics import roc_auc_score\nfrom sklearn.preprocessing import LabelEncoder","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df = pd.read_csv('/kaggle/input/riiid-test-answer-prediction/train.csv',\n                         usecols=['row_id', 'user_id', 'answered_correctly', 'content_id', 'prior_question_had_explanation', 'prior_question_elapsed_time'],\n                         dtype={'row_id': 'int64',  'user_id': 'int32', 'content_id': 'int16', 'answered_correctly': 'int8', 'prior_question_had_explanation': 'boolean', 'prior_question_elapsed_time':'float32'}\n                         )\n\ntrain_df.shape","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"user_df = train_df[train_df.answered_correctly != -1].groupby('user_id').agg({'answered_correctly': ['count', 'mean']}).reset_index()\nuser_df.columns = ['user_id', 'user_questions', 'user_mean']\nuser_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"user_lect = train_df.groupby([\"user_id\", \"answered_correctly\"]).size().unstack()\nuser_lect.columns = ['lecture', 'wrong', 'right']\nuser_lect['lecture'] = user_lect['lecture'].fillna(0)\nuser_lect = user_lect.astype('Int64')\nuser_lect['watches_lecture'] = np.where(user_lect.lecture > 0, 1, 0)\nuser_lect = user_lect.reset_index()\nuser_lect = user_lect[['user_id', 'watches_lecture']]\nuser_lect.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"user_df = user_df.merge(user_lect, on = \"user_id\", how = \"left\")\ndel user_lect\nuser_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"content_df = train_df[train_df.answered_correctly != -1].groupby('content_id').agg({'answered_correctly': ['count', 'mean']}).reset_index()\ncontent_df.columns = ['content_id', 'content_questions', 'content_mean']\ncontent_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"cv2_train = pd.read_pickle(\"../input/riidvalidationpickle/cv2_train.pickle\")['row_id']\ncv2_valid = pd.read_pickle(\"../input/riidvalidationpickle/cv2_valid.pickle\")['row_id']","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"import gc\ntrain_df = train_df[train_df.answered_correctly != -1]\nmean_prior = train_df.prior_question_elapsed_time.astype(\"float64\").mean()\n\nvalidation_df = train_df[train_df.row_id.isin(cv2_valid)]\ntrain_df = train_df[train_df.row_id.isin(cv2_train)]\n\nvalidation_df = validation_df.drop(columns = \"row_id\")\ntrain_df = train_df.drop(columns = \"row_id\")\n\ndel cv2_train, cv2_valid\ngc.collect()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"label_enc = LabelEncoder()\n\ntrain_df= train_df.merge(user_df, on = \"user_id\", how = \"left\")\ntrain_df = train_df.merge(content_df, on = \"content_id\", how = \"left\")\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df['content_questions'].fillna(0, inplace = True)\ntrain_df['content_mean'].fillna(0.5, inplace = True)\ntrain_df['watches_lecture'].fillna(0, inplace = True)\ntrain_df['user_questions'].fillna(0, inplace = True)\ntrain_df['user_mean'].fillna(0.5, inplace = True)\ntrain_df['prior_question_elapsed_time'].fillna(mean_prior, inplace = True)\ntrain_df['prior_question_had_explanation'].fillna(False, inplace = True)\ntrain_df['prior_question_had_explanation'] = label_enc.fit_transform(train_df['prior_question_had_explanation'])\ntrain_df[['content_questions', 'user_questions']] = train_df[['content_questions', 'user_questions']].astype(int)\ntrain_df.sample(5)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"validation_df = validation_df.merge(user_df, on = \"user_id\", how = \"left\")\nvalidation_df = validation_df.merge(content_df, on = \"content_id\", how = \"left\")\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"validation_df['content_questions'].fillna(0, inplace = True)\nvalidation_df['content_mean'].fillna(0.5, inplace = True)\nvalidation_df['watches_lecture'].fillna(0, inplace = True)\nvalidation_df['user_questions'].fillna(0, inplace = True)\nvalidation_df['user_mean'].fillna(0.5, inplace = True)\nvalidation_df['prior_question_had_explanation'].fillna(False, inplace = True)\nvalidation_df['prior_question_elapsed_time'].fillna(mean_prior, inplace = True)\nvalidation_df['prior_question_had_explanation'] = label_enc.fit_transform(validation_df['prior_question_had_explanation'])\nvalidation_df[['content_questions', 'user_questions']] = validation_df[['content_questions', 'user_questions']].astype(int)\nvalidation_df.sample(5)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df.shape","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"features = ['user_questions', 'user_mean', 'content_questions', 'content_mean', \n            'prior_question_had_explanation', 'prior_question_elapsed_time', 'watches_lecture']\n\n\ntrain = train_df.sample(n=5000000, random_state = 1)\n\ny_train = train['answered_correctly']\ntrain = train[features]\n\ny_val = validation_df['answered_correctly']\nvalidation = validation_df[features]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"params = {'objective': 'binary',\n          'metric': 'auc',\n          'seed': 42,\n          'learning_rate': 0.1, \n          \"boosting_type\": \"gbdt\" \n         }","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"lgb_train = lgb.Dataset(train, y_train, categorical_feature = None)\nlgb_eval = lgb.Dataset(validation, y_val, categorical_feature = None)\ndel train, y_train, validation, y_val\ngc.collect()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model = lgb.train(\n    params, lgb_train,\n    valid_sets=[lgb_train, lgb_eval],\n    verbose_eval=50,\n    num_boost_round=10000,\n    early_stopping_rounds=8\n)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"lgb.plot_importance(model)\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"import riiideducation\nenv = riiideducation.make_env()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"iter_test = env.iter_test()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for (test_df, sample_prediction_df) in iter_test:\n    test_df = test_df.merge(user_df, on = \"user_id\", how = \"left\")\n    test_df = test_df.merge(content_df, on = \"content_id\", how = \"left\")\n    test_df['content_questions'].fillna(0, inplace = True)\n    test_df['content_mean'].fillna(0.5, inplace = True)\n    test_df['watches_lecture'].fillna(0, inplace = True)\n    test_df['user_questions'].fillna(0, inplace = True)\n    test_df['user_mean'].fillna(0.5, inplace = True)\n    test_df['prior_question_elapsed_time'].fillna(mean_prior, inplace = True)\n    test_df['prior_question_had_explanation'].fillna(False, inplace = True)\n    test_df['prior_question_had_explanation'] = label_enc.fit_transform(test_df['prior_question_had_explanation'])\n    test_df[['content_questions', 'user_questions']] = test_df[['content_questions', 'user_questions']].astype(int)\n    test_df['answered_correctly'] =  model.predict(test_df[features])\n    env.predict(test_df.loc[test_df['content_type_id'] == 0, ['row_id', 'answered_correctly']])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}