Shekhar/feature_study.py at master · heliohackweek/Shekhar · 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
132
133
134
135
136
137
138
139
140
141
142
143
# -*- coding: utf-8 -*-
"""feature_study.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/14UR7nxS8giDXLEM_NETD2KZ5_2KEhmAM
"""

# imports for the random forest
import numpy as np
import scipy.stats as stats
import pandas as pd
import glob
import csv
import matplotlib.pyplot as plt


import sklearn
from sklearn.linear_model import LinearRegression
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.model_selection import train_test_split
from sklearn.model_selection import cross_val_score
from sklearn.metrics import accuracy_score
from sklearn import metrics


from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
from sklearn.ensemble import RandomForestRegressor

## Load in full data set
# dateTime = 0 ', 'Kp= 1', 'Dst=2', 'dens=3', 'velo=4', 'Pdyn=5', 'ByIMF=6', 'BzIMF=7',
#        'mep06=8', 'mep05', 'mep01', 'me03', 'Lval', 'mlt', 'lat', 'lon'],
maxset = pd.read_csv("/content/csv_data/metop_rad_belt_passes_max.csv")
minset = pd.read_csv("/content/csv_data/metop_rad_belt_passes_min.csv")
medset = pd.read_csv("/content/csv_data/metop_rad_belt_passes_median.csv")

dataset=maxset
dataset.columns = ['maxdateTime', 'maxKp', 'maxDst', 'maxdens', 'maxvelo',
                   'maxPdyn', 'maxByIMF', 'maxBzIMF', 'maxmep06', 'maxmep05',
                   'maxmep01', 'maxme03', 'maxLval', 'maxmlt', 'maxlat',
                   'maxlon']
dataset['MinBz'] = minset.BzIMF
dataset['MinBy'] = minset.ByIMF
dataset['medKp'] = medset.Kp
dataset['medDst'] = medset.Dst
dataset['meddens'] = medset.dens
dataset['medvelo'] = medset.velo
dataset['medPdyn'] = medset.Pdyn
dataset['medmep06'] = medset.mep06
dataset['maxmomentum'] = dataset.maxdens * dataset.maxvelo
dataset['medmomentum'] = dataset.meddens * dataset.medvelo

dropcolumns = ['maxdateTime', 'maxmep05', 'maxmep01', 'maxme03', 'maxLval',
               'maxmlt', 'maxlat', 'maxlon']
dataset = dataset.drop(columns=dropcolumns)

# ['maxKp', 'maxDst', 'maxdens', 'maxvelo', 'maxPdyn', 'maxByIMF', 'maxBzIMF',
#    'maxmep06', 'MinBz', 'MinBy', 'medKp', 'medDst', 'meddens',
#    'medvelo', 'medPdyn', 'medmep06', 'maxmomentum', 'medmomentum']

#moredrops = [0,1,2]
#dataset = dataset.drop(columns = dataset.columns.values[moredrops])

# loading in the data and separating out the x and y
# and the train and test set

testfrac = .8  # fraction of set to use for training
boolval = 1    # 0 for linear regression, 1 for boolean threshold
boolthresh = .8  # if boolean regression, set threshold of mep06 to count as an event
shufval = 0    # set 1 to separate testing set through shuffling, 0 to window
trainmed = 1  # if 0, train on the max of mep06, if 1, train on the median instead
estimators = 300  # number of estimators in random forest
setlen = len(dataset)

if (shufval):
  train_dataset = dataset.sample(frac=testfrac,random_state=0)
  test_dataset = dataset.drop(train_dataset.index)
else:
  train_dataset = dataset[:round(testfrac*setlen)]
  test_dataset = dataset[round(testfrac*setlen):]

if (trainmed):
  y_train = train_dataset.pop('medmep06')
  y_test = test_dataset.pop('medmep06')
  x_train = train_dataset.drop(columns = 'maxmep06')
  x_test = test_dataset.drop(columns = 'maxmep06')
else:
  y_train = train_dataset.pop('maxmep06')
  y_test = test_dataset.pop('maxmep06')
  x_train = train_dataset.drop(columns = 'medmep06')
  x_test = test_dataset.drop(columns = 'medmep06')

if (boolval):
  y_train = (y_train>boolthresh)*1.
  y_test = (y_test>boolthresh)*1.

regressor = RandomForestRegressor(n_estimators = estimators,  oob_score = True, random_state = 0)
regressor.fit(x_train, y_train)
Y_pred = regressor.predict(x_test)
if (boolval):
  Y_pred_bool = (Y_pred>boolthresh)*1.
mae=metrics.mean_absolute_error(y_test, Y_pred_bool)
mse=metrics.mean_squared_error(y_test, Y_pred_bool)
trainsc = regressor.score(x_train, y_train)
testsc = regressor.score(x_test, y_test)
regoob = regressor.oob_score_
importances = regressor.feature_importances_
std = np.std([tree.feature_importances_ for tree in regressor.estimators_],
             axis=0)
indices = np.argsort(importances)[::-1]

## Evaluate performance
print("mean absolute error: %f;  mean squared error: %f" % (mae, mse))
print('R^2 Training Score: {:.2f} \nOOB Score: {:.2f} \nR^2 Test Score: {:.2f}'.format(regressor.score(x_train, y_train),                                                                                   regressor.oob_score_,
                                                                                             regressor.score(x_test, y_test)))

print("Feature ranking:")

for f in range(x_train.shape[1]):
    print("%d. feature %s (%f)" % (f + 1, x_test.columns.values[round(f)], importances[indices[f]]))

# Plot the impurity-based feature importances of the forest
plt.figure()
plt.title("Feature importances")
plt.bar(range(x_train.shape[1]), importances[indices],
        color="r", yerr=std[indices], align="center")
plt.xticks(range(x_train.shape[1]), indices)
plt.xlim([-1, x_train.shape[1]])
plt.show()

cm = sklearn.metrics.confusion_matrix(y_test, Y_pred_bool).ravel()
# true negatives, false positives, false negatives, true positives
print(cm)

# set up evaluation table
# parameters are [shufval, testfrac, boolval, boolthresh, setlen, trainmed, estimators]
params = np.array([shufval, testfrac, boolval, boolthresh, setlen, trainmed, estimators])
eval_table = pd.DataFrame([params, params, params, params, params])
eval_table.shape
eval_table.index = ['mae', 'mse', 'r2train', 'r2test', 'oob']
eval_table.columns = ['shufval', 'testfrac', 'boolval', 'boolthresh', 'setlen', 'trainmed', 'estimators']