{"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_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import pandas as pd, numpy as np, gc\nfrom sklearn.model_selection import KFold, GroupKFold\nfrom xgboost import XGBClassifier\nfrom sklearn.metrics import f1_score","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Load Train Data and Labels\nTo avoid memory error, we will read the train data in as 10 pieces and feature engineer each piece before reading the next piece. ","metadata":{}},{"cell_type":"code","source":"# READ USER ID ONLY\ntmp = pd.read_csv(\"/kaggle/input/predict-student-performance-from-game-play/train.csv\",usecols=[0])\ntmp = tmp.groupby('session_id').session_id.agg('count')\n\n# COMPUTE READS AND SKIPS\nPIECES = 10\nCHUNK = int( np.ceil(len(tmp)/PIECES) )\n\nreads = []\nskips = [0]\nfor k in range(PIECES):\n    a = k*CHUNK\n    b = (k+1)*CHUNK\n    if b>len(tmp): b=len(tmp)\n    r = tmp.iloc[a:b].sum()\n    reads.append(r)\n    skips.append(skips[-1]+r)\n    \nprint(f'To avoid memory error, we will read train in {PIECES} pieces of sizes:')\nprint(reads)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train = pd.read_csv('/kaggle/input/predict-student-performance-from-game-play/train.csv', nrows=reads[0])\nprint('Train size of first piece:', train.shape )\ntrain.head()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"targets = pd.read_csv('/kaggle/input/predict-student-performance-from-game-play/train_labels.csv')\ntargets['session'] = targets.session_id.apply(lambda x: int(x.split('_')[0]) )\ntargets['q'] = targets.session_id.apply(lambda x: int(x.split('_')[-1][1:]) )\nprint( targets.shape )\ntargets.head()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Feature Engineer\nWe create basic aggregate features.","metadata":{}},{"cell_type":"code","source":"CATS = ['event_name', 'fqid', 'room_fqid', 'text']\nNUMS = ['elapsed_time','level','page','room_coor_x', 'room_coor_y', \n        'screen_coor_x', 'screen_coor_y', 'hover_duration']\n\nEVENTS = ['navigate_click','person_click','cutscene_click','object_click',\n          'map_hover','notification_click','map_click','observation_click',\n          'checkpoint']","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def feature_engineer(train):\n    \n    dfs = []\n    for c in CATS:\n        tmp = train.groupby(['session_id','level_group'])[c].agg('nunique')\n        tmp.name = tmp.name + '_nunique'\n        dfs.append(tmp)\n    for c in NUMS:\n        tmp = train.groupby(['session_id','level_group'])[c].agg('mean')\n        tmp.name = tmp.name + '_mean'\n        dfs.append(tmp)\n    for c in NUMS:\n        tmp = train.groupby(['session_id','level_group'])[c].agg('std')\n        tmp.name = tmp.name + '_std'\n        dfs.append(tmp)\n    for c in EVENTS: \n        train[c] = (train.event_name == c).astype('int8')\n    for c in EVENTS + ['elapsed_time']:\n        tmp = train.groupby(['session_id','level_group'])[c].agg('sum')\n        tmp.name = tmp.name + '_sum'\n        dfs.append(tmp)\n    train = train.drop(EVENTS,axis=1)\n        \n    df = pd.concat(dfs,axis=1)\n    df = df.fillna(-1)\n    df = df.reset_index()\n    df = df.set_index('session_id')\n    return df","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\n\n# PROCESS TRAIN DATA IN PIECES\nall_pieces = []\nprint(f'Processing train as {PIECES} pieces to avoid memory error... ')\nfor k in range(PIECES):\n    print(k,', ',end='')\n    SKIPS = 0\n    if k>0: SKIPS = range(1,skips[k]+1)\n    train = pd.read_csv('/kaggle/input/predict-student-performance-from-game-play/train.csv',\n                        nrows=reads[k], skiprows=SKIPS)\n    df = feature_engineer(train)\n    all_pieces.append(df)\n    \n# CONCATENATE ALL PIECES\nprint('\\n')\ndel train; gc.collect()\ndf = pd.concat(all_pieces, axis=0)\nprint('Shape of all train data after feature engineering:', df.shape )\ndf.head()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Train XGBoost Model\nWe train one model for each of 18 questions. Furthermore, we use data from `level_groups = '0-4'` to train model for questions 1-3, and `level groups '5-12'` to train questions 4 thru 13 and `level groups '13-22'` to train questions 14 thru 18. Because this is the data we get (to predict corresponding questions) from Kaggle's inference API during test inference. We can improve our model by saving a user's previous data from earlier `level_groups` and using that to predict future `level_groups`.","metadata":{}},{"cell_type":"code","source":"FEATURES = [c for c in df.columns if c != 'level_group']\nprint('We will train with', len(FEATURES) ,'features')\nALL_USERS = df.index.unique()\nprint('We will train with', len(ALL_USERS) ,'users info')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"gkf = GroupKFold(n_splits=5)\noof = pd.DataFrame(data=np.zeros((len(ALL_USERS),18)), index=ALL_USERS)\nmodels = {}\n\n# COMPUTE CV SCORE WITH 5 GROUP K FOLD\nfor i, (train_index, test_index) in enumerate(gkf.split(X=df, groups=df.index)):\n    print('#'*25)\n    print('### Fold',i+1)\n    print('#'*25)\n    \n    xgb_params = {\n    'objective' : 'binary:logistic',\n    'eval_metric':'logloss',\n    'learning_rate': 0.05,\n    'max_depth': 4,\n    'n_estimators': 1000,\n    'early_stopping_rounds': 50,\n    'tree_method':'hist',\n    'subsample':0.8,\n    'colsample_bytree': 0.4,\n    'use_label_encoder' : False}\n    \n    # ITERATE THRU QUESTIONS 1 THRU 18\n    for t in range(1,19):\n        \n        # USE THIS TRAIN DATA WITH THESE QUESTIONS\n        if t<=3: grp = '0-4'\n        elif t<=13: grp = '5-12'\n        elif t<=22: grp = '13-22'\n            \n        # TRAIN DATA\n        train_x = df.iloc[train_index]\n        train_x = train_x.loc[train_x.level_group == grp]\n        train_users = train_x.index.values\n        train_y = targets.loc[targets.q==t].set_index('session').loc[train_users]\n        \n        # VALID DATA\n        valid_x = df.iloc[test_index]\n        valid_x = valid_x.loc[valid_x.level_group == grp]\n        valid_users = valid_x.index.values\n        valid_y = targets.loc[targets.q==t].set_index('session').loc[valid_users]\n        \n        # TRAIN MODEL        \n        clf =  XGBClassifier(**xgb_params)\n        clf.fit(train_x[FEATURES].astype('float32'), train_y['correct'],\n                eval_set=[ (valid_x[FEATURES].astype('float32'), valid_y['correct']) ],\n                verbose=0)\n        print(f'{t}({clf.best_ntree_limit}), ',end='')\n        \n        # SAVE MODEL, PREDICT VALID OOF\n        models[f'{grp}_{t}'] = clf\n        oof.loc[valid_users, t-1] = clf.predict_proba(valid_x[FEATURES].astype('float32'))[:,1]\n        \n    print()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# PUT TRUE LABELS INTO DATAFRAME WITH 18 COLUMNS\ntrue = oof.copy()\nfor k in range(18):\n    # GET TRUE LABELS\n    tmp = targets.loc[targets.q == k+1].set_index('session').loc[ALL_USERS]\n    true[k] = tmp.correct.values","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# FIND BEST THRESHOLD TO CONVERT PROBS INTO 1s AND 0s\nscores = []; thresholds = []\nbest_score = 0; best_threshold = 0\n\nfor threshold in np.arange(0.4,0.81,0.01):\n    print(f'{threshold:.02f}, ',end='')\n    preds = (oof.values.reshape((-1))>threshold).astype('int')\n    m = f1_score(true.values.reshape((-1)), preds, average='macro')   \n    scores.append(m)\n    thresholds.append(threshold)\n    if m>best_score:\n        best_score = m\n        best_threshold = threshold","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import matplotlib.pyplot as plt\n\n# PLOT THRESHOLD VS. F1_SCORE\nplt.figure(figsize=(20,5))\nplt.plot(thresholds,scores,'-o',color='blue')\nplt.scatter([best_threshold], [best_score], color='blue', s=300, alpha=1)\nplt.xlabel('Threshold',size=14)\nplt.ylabel('Validation F1 Score',size=14)\nplt.title(f'Threshold vs. F1_Score with Best F1_Score = {best_score:.3f} at Best Threshold = {best_threshold:.3}',size=18)\nplt.show()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print('When using optimal threshold...')\nfor k in range(18):\n        \n    # COMPUTE F1 SCORE PER QUESTION\n    m = f1_score(true[k].values, (oof[k].values>best_threshold).astype('int'), average='macro')\n    print(f'Q{k}: F1 =',m)\n    \n# COMPUTE F1 SCORE OVERALL\nm = f1_score(true.values.reshape((-1)), (oof.values.reshape((-1))>best_threshold).astype('int'), average='macro')\nprint('==> Overall F1 =',m)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# IMPORT KAGGLE API\nimport jo_wilder\nenv = jo_wilder.make_env()\niter_test = env.iter_test()\n\n# CLEAR MEMORY\nimport gc\ndel targets, df, oof, true\n_ = gc.collect()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"limits = {'0-4':(1,4), '5-12':(4,14), '13-22':(14,19)}\n\nfor (test, sample_submission) in iter_test:\n    \n    # FEATURE ENGINEER TEST DATA\n    df = feature_engineer(test)\n    \n    # INFER TEST DATA\n    grp = test.level_group.values[0]\n    a,b = limits[grp]\n    for t in range(a,b):\n        clf = models[f'{grp}_{t}']\n        p = clf.predict_proba(df[FEATURES].astype('float32'))[0,1]\n        mask = sample_submission.session_id.str.contains(f'q{t}')\n        sample_submission.loc[mask,'correct'] = int( p > best_threshold )\n    \n    env.predict(sample_submission)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df = pd.read_csv('submission.csv')\nprint( df.shape )\ndf.head()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(df.correct.mean())","metadata":{},"execution_count":null,"outputs":[]}]}