ln2sql/model_builder.py at master · karthikrab/ln2sql · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from tpot import TPOTClassifier, TPOTRegressor
from sklearn.metrics.scorer import make_scorer
import dill as pickle
import os
import json

col_one_hot={}
col_classes={}

def my_custom_accuracy(y_true, y_pred):
    return float(sum(y_pred == y_true)) / len(y_true)


def train_tpot_classifier(dat, target,tname):
    # print(target)
    dat.to_csv('tpot_inp_'+tname+"_"+ str(target) + '.csv')
    df = dat.rename(columns={target: 'class'})
    X = df.drop(columns='class')
    y = df['class'].copy()
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)
    my_custom_scorer = make_scorer(my_custom_accuracy, greater_is_better=True)
    # print('X_train)
    tpot = TPOTClassifier(generations=1, population_size=20, verbosity=2,
                          scoring=my_custom_scorer)
    tpot.fit(X_train, y_train)
    print(tpot.score(X_test, y_test))
    tpot.export('tpot_pipeline_' +tname+"_"+ str(target) + '.py')
    os.system('/home/dev/T2DB/ln2sql/fixtpot.sh tpot_pipeline_' +tname+"_"+ str(target) + '.py '+str(target))
    os.system('python3 tpot_pipeline_' +tname+"_"+ str(target) + '.py ' 'tpot_inp_'+tname+"_"+ str(target) + '.csv')

def train_tpot_regressor(dat, target):
    # print(target)
    df = dat.rename(columns={target: 'class'})
    X = df.drop(columns='class')
    y = df['class'].copy()
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)
    my_custom_scorer = make_scorer(my_custom_accuracy, greater_is_better=True)

    tpot = TPOTRegressor(generations=1, population_size=20, verbosity=2,
                         scoring=my_custom_scorer)
    tpot.fit(X_train, y_train)
    #tpot.export('tpot_mnist_pipeline_'+tname+"_"+ str(target) + '.py')
    with open('tpot_pkl_' +tname+"_"+ str(target) + '.pkl','wb') as f:
        pickle.dump(tpot,f)

def one_hot(dat, columns):
    final_dat = pd.DataFrame()
    for column_name in columns:
        dat_col = pd.get_dummies(dat[column_name], prefix=column_name, drop_first=True)
        final_dat = pd.concat([final_dat, dat_col], axis=1, sort=True)
        col_one_hot[column_name]=list(dat_col.columns)
    print(col_one_hot)
    return final_dat


def model_builder(filename,tname):
    data = pd.read_csv(filename)
    string_colums = []
    numeric_columns = []
    float_column = []
    column_list = []
    misc_columns = []
    numeric_list = []
    float_list = []
    for i, v in data.dtypes.items():
        if v == np.dtype(object):
            d_type = i, 'String'
            string_colums.append(d_type)
        elif np.issubdtype(v, int):
            d_type = i, 'Integer'
            numeric_columns.append(d_type)
        elif v == np.dtype(float):
            d_type = i, 'Float'
            float_column.append(d_type)
        else:
            d_type = i, v
            misc_columns.append(d_type)

    for i, v in string_colums:
        if data[i].nunique() < 20:
            column_list.append(i)
            col_classes[i]=list(data[i].unique())

    for i, v in numeric_columns:
        if data[i].nunique() <= 4:
            numeric_list.append(i)
            col_classes[i]=list(data[i].unique())
    with open('tpot_class_' +tname + '.json','w')as f:
        json.dump(col_classes,f)
    for i, v in float_column:
        float_list.append(i)

    for items in column_list:
        iters = 0
        target = data[items].copy()
        one_hot_list = [x for x in column_list if x != items]
        depend_data = data.drop(columns=items)
        one_hot_data = one_hot(depend_data, one_hot_list)
        final_data = pd.concat([one_hot_data, target], axis=1, sort=True)
        final_data = pd.concat([final_data, data.drop(columns=column_list)], axis=1, sort=True)
        for i in data[items].unique():
            final_data[items] = final_data[items].mask(final_data[items] == i, iters)
            iters += 1
        if target.dtype == np.dtype(object):
            final_data[items] = final_data[items].astype('category')
        train_tpot_classifier(final_data, items,tname)

    for items in numeric_list:
        target = data[items].copy()
        one_hot_data = one_hot(data[column_list], column_list)
        final_data = pd.concat([one_hot_data, target], axis=1, sort=True)
        depend_data = data.drop(columns=column_list, axis=1)
        depend_data = depend_data.drop(columns=items, axis=1)
        final_data = pd.concat([final_data, depend_data], axis=1, sort=True)
        train_tpot_classifier(final_data, items,tname)

    for items in float_list:
        target = data[items].copy()
        one_hot_data = one_hot(data[column_list], column_list)
        final_data = pd.concat([one_hot_data, target], axis=1, sort=True)
        depend_data = data.drop(columns=column_list, axis=1)
        depend_data = depend_data.drop(columns=items, axis=1)
        final_data = pd.concat([final_data, depend_data], axis=1, sort=True)
        train_tpot_regressor(final_data, items,tname)


if __name__ == "__main__":
    model_builder('bank.csv')