RovOpt/rov_train_classify.py

import warnings

warnings.filterwarnings("ignore")
import os
import pandas as pd
import gc
import math
import numpy as np
import time
import lightgbm as lgb
from sklearn.model_selection import train_test_split
from sklearn.model_selection import StratifiedKFold
from sklearn import metrics
import pickle
from sklearn.metrics import top_k_accuracy_score
import seaborn as sns
import matplotlib.pylab as plt
from odps import ODPS
from odps.df import DataFrame as odpsdf
from datetime import datetime as dt
import datetime
import process_feature
import _pickle as cPickle
from sklearn.feature_selection import SelectFromModel
from sklearn.linear_model import LogisticRegression

def getRovfeaturetable(dt, table):
    odps = ODPS('LTAI4FtW5ZzxMvdw35aNkmcp', '0VKnydcaHK3ITjylbgUsLubX6rwiwc', 'usercdm',
                endpoint='http://service.cn.maxcompute.aliyun.com/api', connect_timeout=3000, \
                read_timeout=500000, pool_maxsize=1000, pool_connections=1000)

    featureArray = []
    for record in odps.read_table(table, partition='dt=%s' % dt):
        valueFeature = {}
        for i in process_feature.featurename:
            if i == 'dt':
                valueFeature[i] = dt
            else:
                valueFeature[i] = record[i]
        featureArray.append(valueFeature)
    featureArray = pd.DataFrame(featureArray)
    print(dt, table, 'feature table finish')
    return featureArray

def getdatasample(date, max_range, table):
    new_date = dt.strptime(date, '%Y%m%d')
    datelist = []
    testlist = []
    for i in range(0, max_range):
        delta = datetime.timedelta(days=i)
        tar_dt = new_date - delta
        datelist.append(tar_dt.strftime("%Y%m%d"))
    for tm in datelist:
        testlist.append(getRovfeaturetable(tm, table))
    data = pd.concat(testlist)
    data.reset_index(inplace=True)
    data = data.drop(axis=1, columns='index')
    return data

def discrete_y(y):
    y = float(y)
    if y>1000000:
        return 7
    elif y>500000:
        return 6
    elif y>100000:
        return 5
    elif y>50000:
        return 4
    elif y>10000:
        return 3
    elif y>5000:
        return 2
    elif y>1000:
        return 1
    else:
        return 0

def clean_data(df):
    #y = df['futre7dayreturn'].apply(lambda x: np.log(df['futre7dayreturn']+1))
    df['futre7dayreturn'] = df['futre7dayreturn'].apply(discrete_y)
    y = df['futre7dayreturn']
    print(y)
    df_vids = df['videoid']
    #drop string
    #x = df.drop(['videoid', 'videotags', 'videotitle', 'videodescr', 'videodistribute_title', 'videoallwords', 'words_without_tags'], axis=1)
    x = df.drop(['videoid', 'videotags', 'words_without_tags', 'dt'], axis=1)
    #drop future
    #x = df.drop(['futr5viewcount', 'futr5returncount', 'futre7dayreturn'], axis=1)
    x = x.drop(['futre7dayreturn'], axis=1)

    x['stage_four_retrn_added'] = x['stage_four_retrn'] - x['stage_three_retrn']
    x['stage_three_retrn_added'] = x['stage_three_retrn'] - x['stage_two_retrn']
    x['stage_two_retrn_added'] = x['stage_two_retrn'] - x['stage_one_retrn']

    x['stage_four_retrn_ratio'] = (x['stage_four_retrn'] - x['stage_three_retrn'])/x['stage_four_retrn']
    x['stage_three_retrn_ratio'] = (x['stage_three_retrn'] - x['stage_two_retrn'])/x['stage_three_retrn']
    x['stage_two_retrn_ratio'] = (x['stage_two_retrn'] - x['stage_one_retrn'])/x['stage_two_retrn']

    features = list(x)
    drop_features = [f for f in features if (f.find('day30')!=-1 or f.find('day60')!=-1)]
    x = x.drop(drop_features, axis=1)
    x = x.fillna(0)
    x = x.astype('float64')
    #x.fillna(0)
    x.clip(0,2000000)
 
    features = [f for f in features if f not in drop_features]
    return x, y , df_vids, features

def feature_selection(X, y):
    selector = SelectFromModel(estimator=LogisticRegression()).fit(X, y)
    return selector

def auto_train(X_train, y_train):
    from flaml import  AutoML
    automl = AutoML()
    automl_settings = {
    "time_budget": 10,  # in seconds
    "metric": 'r2',
    "task": 'regression',
    "log_file_name": "test/auto.log",
    "estimator_list": ["lgbm"]
    }
    automl.fit(X_train=X_train, y_train=y_train,
           **automl_settings) 

    pred_test_y = automl.predict(X_train)
    y_test = y_train.values

    #err_mape = mean_absolute_percentage_error(y_test, pred_test_y)
    r2 = r2_score(y_test, pred_test_y)
    #print('err_mape', err_mape)
    print('r2', r2)

    pack_result(pred_test_y, y_test,[],'autoval.csv')

    

def train(x,y,features):
    X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.33, stratify=y )

    '''
    selector = feature_selection(X_train, y_train) 
    X_train = selector.transform(X_train)
    X_test = selector.transform(X_test)
    selected_features = []
    _supported = selector.get_support()
    for i in range(0, len(_supported)):
        if _supported[i]:
            selected_features.append(features[i])
    features = selected_features 
    '''

    print(len(X_train), len(X_test))
    params = {
        "objective": "multiclass",
        "num_classes": 8,
        "max_depth": 6,
        "num_leaves": 30,
        "learning_rate": 0.05,
        "bagging_fraction": 0.7,
        "feature_fraction": 0.7,
        "bagging_freq": 8,
        "bagging_seed": 2018,
        "lambda_l1": 0.1,
        "boosting": "gbdt",
        "nthread": 4,
        "verbosity": -1
    }
    lgtrain = lgb.Dataset(X_train, label=y_train)
    lgval = lgb.Dataset(X_test, label=y_test)
    evals_result = {}
    model = lgb.train(params, lgtrain, 10000, valid_sets=[lgval], early_stopping_rounds=200, verbose_eval=20,
                      evals_result=evals_result)

    pack_result(model.feature_importance(), features, [], 'importance.csv')
    
    pred_test_y = model.predict(X_test, num_iteration=model.best_iteration)
    
    top_k_accuracy_score(y_test, pred_test_y, k=1)
    print('top_k_accuracy_score', top_k_accuracy_score) 

    pack_result(pred_test_y, y_test,[],'val.csv')

    return pred_test_y, model, evals_result


def pack_result(y_, y, vid, fp):
    #y_ = y_.astype(int)
    y_.reshape(len(y_),1) 
    df = pd.DataFrame(data=y_, columns=['score'])
    if len(vid) >0:
        df['vid'] = vid
    df['y'] = y
    df = df.sort_values(by=['score'], ascending=False)
    df.to_csv(fp, index=False)

    
if __name__ == '__main__':
    with open(r"train_data_x.pickle", "rb") as input_file:
        train_data = cPickle.load(input_file)    
    with open(r"predict_data_x.pickle", "rb") as input_file:
        predict_data = cPickle.load(input_file)   

    x,y,_,features = clean_data(train_data)

    #auto train
    #auto_train(x,y)

    #train
    _, model, _ = train(x, y, features)
    with open('model.pickle','wb') as output_file:
        cPickle.dump(model, output_file)
    '''
    with open(r"model.pickle", "rb") as input_file:
        model = cPickle.load(input_file)    
    ''' 
    x,y,vid,_ = clean_data(predict_data)
    y_ = model.predict(x, num_iteration=model.best_iteration)

    pack_result(y_, y, vid, 'pred.csv')