-
Notifications
You must be signed in to change notification settings - Fork 0
/
regression.py
142 lines (112 loc) · 4.05 KB
/
regression.py
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
132
133
134
135
136
137
138
139
140
141
142
from turtle import color
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn import datasets, linear_model, model_selection, metrics, ensemble, svm, neighbors
def estimator(X, Y):
seed = 10
# Train, Validation and Test Split
X_train, X_val, Y_train, Y_val = model_selection.train_test_split(X, Y,test_size=0.20, random_state=seed)
X_val, X_test, Y_val, Y_test = model_selection.train_test_split(X_val, Y_val,test_size=0.50, random_state=seed)
print(X_train.shape)
print(X_val.shape)
print(X_test.shape)
def evaluate(model):
# Get predictions by the model
Y_pred_train = model.predict(X_train)
Y_pred_val = model.predict(X_val)
# Compute errors
train_errors = Y_train - Y_pred_train
val_errors = Y_val - Y_pred_val
# Compute MSE
train_mse = metrics.mean_squared_error(Y_train, Y_pred_train)
val_mse = metrics.mean_squared_error(Y_val, Y_pred_val)
# Plot errors
plt.subplot(1, 2, 1)
plt.hist(train_errors, color='blue')
plt.title('Train Error')
plt.subplot(1, 2, 2)
plt.hist(val_errors, color='red')
plt.title('Validation Error')
plt.suptitle(f'{model.__class__}')
plt.show()
# Print results
print(f'{model.__class__}:')
print(f'\tTraining MSE\t: {train_mse}')
print(f'\tValidation MSE\t: {val_mse}', '\n\n')
return val_mse
trained_models = []
scores = []
# Linear Regression
lr_model = linear_model.LinearRegression()
lr_model.fit(X_train, Y_train)
trained_models.append(lr_model)
scores.append(evaluate(lr_model))
# Bagging Regression
bag_model = ensemble.BaggingRegressor(random_state=seed)
bag_model.fit(X_train, Y_train)
trained_models.append(bag_model)
scores.append(evaluate(bag_model))
# Random Forest Regression
rf_model = ensemble.RandomForestRegressor(random_state=seed)
rf_model.fit(X_train, Y_train)
trained_models.append(rf_model)
scores.append(evaluate(rf_model))
# SVM Regression
svm_model = svm.LinearSVR(random_state=seed)
svm_model.fit(X_train, Y_train)
trained_models.append(svm_model)
scores.append(evaluate(svm_model))
# KNN Regression
knn_model = neighbors.KNeighborsRegressor()
knn_model.fit(X_train, Y_train)
trained_models.append(knn_model)
scores.append(evaluate(knn_model))
# Choose the best model (least mse score)
best_model = trained_models[scores.index(min(scores))]
Y_pred_test = best_model.predict(X_test)
errors = Y_test - Y_pred_test
test_mse = metrics.mean_squared_error(Y_test, Y_pred_test)
# Plot test error
plt.subplot(1, 1, 1)
plt.hist(errors, color='purple')
plt.title('Test Error')
plt.suptitle(f'{best_model.__class__}')
plt.show()
return best_model, test_mse
if __name__ == '__main__':
(X, Y) = datasets.fetch_california_housing(return_X_y=True, as_frame=True)
(m, n) = X.shape
cols = X.columns
print(X.shape, type(X))
print(Y.shape, type(Y))
print(X.info())
print(X.describe())
print(X.isna().sum())
def plot_data():
for i in range(n):
plt.subplot(1, 2, 1)
plt.scatter(X[cols[i]], Y)
plt.xlabel(cols[i])
plt.ylabel('Price')
plt.subplot(1, 2, 2)
# plt.hist(X[cols[i]])
plt.violinplot(X[cols[i]])
plt.xlabel(cols[i])
plt.show()
# plot original data
plot_data()
# Add new feature
X['Loc'] = X['Latitude'] * X['Longitude']
# Log normalization
logcols = ['AveRooms', 'AveBedrms', 'Population', 'AveOccup']
X[logcols] = X[logcols].apply(lambda x: np.log(x))
# Mean normalization and feature scaling
normcols = list(set(list(cols)) - set(logcols))
X[normcols] = X[normcols].apply(lambda x:((x - x.mean())/x.std()))
# plot normalized data
plot_data()
model, mse = estimator(X, Y)
print(f'Best model: {model.__class__}')
print(f'Test MSE: {mse}')