shaan-shah

shaan-shah

shaan-shah

shaan-shah

shaan-shah

shaan-shah

@start_new_thread
def send_mail_final(target_variable,df_train,df_test,email_id,message,date_column=None):

    try:
        df_pred=auto_predictor(Target_Variable=target_variable,data_raw=df_train,n_valid=int(0.1*len(df_train)),data_to_predict=df_test,date_column=date_column)
        df_pred=pd.DataFrame(df_pred)
        df_pred.to_csv('predictions.csv')
        recepient = email_id
        email=EmailMessage('Processed Data',message,EMAIL_HOST_USER,[recepient])
        email.attach_file('predictions.csv')

shaan-shah

def auto_predictor(Target_Variable,data_raw,n_valid,data_to_predict,date_column=None):
    if date_column:

        data_raw['{}'.format(date_column)]= pd.to_datetime(data_raw['{}'.format(date_column)])
        data_to_predict['{}'.format(date_column)]= pd.to_datetime(data_to_predict['{}'.format(date_column)])
        intermed=data_trainer(Target_Variable=Target_Variable,data_raw=data_raw,n_valid=n_valid,date_column=date_column)
        return(auto_applyer(leaf_value=intermed[0],feature_value=intermed[1],feature_list=intermed[2],df_raw1=data_raw,df_test=data_to_predict,target_column=Target_Variable,date_column=date_column))
    else:
        intermed=data_trainer(Target_Variable=Target_Variable,data_raw=data_raw,n_valid=n_valid)
        return(auto_applyer(leaf_value=intermed[0],feature_value=intermed[1],feature_list=intermed[2],df_raw1=data_raw,df_test=data_to_predict,target_column=Target_Variable))

shaan-shah

def auto_applyer(leaf_value,feature_value,feature_list,df_raw1,df_test,target_column,date_column=None):
    reset_rf_samples()

    if date_column:
        if date_column in df_test:
            add_datepart(df_test,date_column)
        if date_column in df_raw1:
            add_datepart(df_raw1,date_column)
    '''First we will pre process both test and raw data'''
    train_cats(df_raw1)

shaan-shah

 ''' from here we are doing the feature engineering'''
    print(min_leaf_a)
    reset_rf_samples()
    z=RandomForestRegressor(n_jobs=-1,min_samples_leaf= min_leaf_a,max_features= max_feature_a,oob_score=False,n_estimators=40)
    z.fit(X_train,y_train)
    fi=rf_feat_importance(z,df)
    score=0
    final_feature_importance_value=0
    feature_importance_value_list=[0,0.001,0.002,0.0025,0.003,0.0035]
    for feature_importance_value in feature_importance_value_list:

shaan-shah

def data_trainer(Target_Variable,data_raw,n_valid,date_column=None):
    df_raw=data_raw
    reset_rf_samples()
    ''' This if statement is to reduce the date part'''
    if date_column:
        add_datepart(df_raw,date_column)

    train_cats(df_raw)
    df,y,nas=proc_df(df_raw,Target_Variable)
    n_trn=len(df)-n_valid