Edith/regressionpart.py at master · F3licity/Edith · 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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
#~~~~~~~~~~~~~~~~~REGRESSION~~~~~~~~~~~~~~~~~
start_time = time.time()
'''
##### Define X and y
X = featureArray[:,0:-1] # features
print "X", X.shape
# print X
y = featureArray[:,-1] # Initiation Rates
print "y", y.shape
# print y
'''
#Do the regression per location
X = []
y = []

for c in range(1,18): #1 to 17 chromosomes.
    allcChromosomes = featureArray[featureArray[:,0] == float(c)] # np.where(featureArray[:,1]==0) Take all the rows per chromosome

    X = allcChromosomes[:,1:-1] #Features
    y = allcChromosomes[:,-1] #initiation rates
    # print "X", X.shape
    # print "y", y.shape


    ##### Calculate the Correlation among the features and the target

    pearsonsCorrelations = []
    spearmanCorrelations = []
    # transposedX = X.T
    if len(y) != 0:
        for i in range(0, len(X.T)):
            pC = scipy.stats.pearsonr(X[:,i], y)
            pearsonsCorrelations.append(pC)
            # print pC
            sC = scipy.stats.spearmanr(X[:,i], y)
            spearmanCorrelations.append(sC)
            # print sC
        # pCorrelation = scipy.stats.pearsonr(X[:,0], y)
        # print pCorrelation
        # print "pearsonsCorrelations", pearsonsCorrelations
        # print "spearmanCorrelations", spearmanCorrelations

        pearsonsCorrelations = np.array(pearsonsCorrelations)
        spearmanCorrelations = np.array(spearmanCorrelations)
        # print "pearsonsCorrelations", pearsonsCorrelations.shape
        # print "spearmanCorrelations", spearmanCorrelations.shape

        h = pearsonsCorrelations[:,0] # Pearson correlation coefficient
        z = spearmanCorrelations[:,0] # Spearman
        x = np.arange(len(h)) # Features

        plt.figure()
        plt.plot(x,h,'g', label='pearsonsCorrelations')
        plt.plot(x,z,'r', label='spearmanCorrelations')
        plt.xlabel("Features")
        plt.ylabel("Correlations")
        plt.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3,
                   ncol=2, mode="expand", borderaxespad=0.)
        plt.grid(True)
        # plt.show()
        plt.savefig("CorrelationCoefficientFor" + `c`+"Chrom"+ `csvFile` + ".png")
        plt.clf()


        # Do SVR
        skf = cross_validation.KFold(len(y),n_folds=5)
        allScores = []
        print "X and Y", X.shape, y.shape
        fold = 1
        for train_index, test_index in skf:
            #print("TRAIN:", train_index, "TEST:", test_index)
            X_train, X_test = X[train_index], X[test_index]
            y_train, y_test = y[train_index], y[test_index]

            svr_linear = svm.SVR(kernel='linear', degree=4, C=1, cache_size=300) #SupportVectorMachine - Support Vector Regression  #RandomForestRegressor(n_estimators=100)
            print X_train.shape
            print y_train.shape
            start_training_time = time.time()
            print "start training"
            svr_linear.fit(X_train, y_train)
            print "Done training in ", time.time() - start_training_time
            score = svr_linear.score(X_test, y_test)
            allScores.append(score)
            y_linear_predictTest = svr_linear.predict(X_test)
            y_linear_predictTrain = svr_linear.predict(X_train)


            #Images
            plt.close('all')
            plt.figure()
            plt.subplots_adjust(hspace=.5, wspace=.5)
            plt.subplot(1,2,1)
            plt.title('SVR train')
            allSampleIndexes=np.arange(len(X_train[:])) #we put the len so to give to each sample a number
            plt.scatter(allSampleIndexes, y_train, c='g', edgecolor='none', s=1, label='train')
            plt.hold('on')
            plt.scatter(allSampleIndexes, y_linear_predictTrain, c='r', edgecolor='none', s=1, label='prediction')
            plt.xlabel('#of rows')
            plt.ylabel('initiation rates (log)')

            plt.subplot(1,2,2)
            allSampleIndexes=np.arange(len(X_test[:])) #we put the len so to give to each sample a number
            plt.scatter(allSampleIndexes, y_test, c='g', edgecolor='none', s=1, label='test')
            plt.hold('on')
            plt.scatter(allSampleIndexes, y_linear_predictTest, c='r', edgecolor='none', s=1, label='prediction')
            plt.xlabel('#of rows')
            plt.ylabel('initiation rates (log)')
            plt.title('SVR test')
            plt.legend()
            plt.suptitle('SVR_AllFeats'+ `csvFile` +`fold`)
            plt.savefig('correctciandrinidata/SVR_AllFeatsFor' + `c` + 'Chrom' + `csvFile` +`fold`+'.svg')
            # plt.show()
            plt.clf()
            fold += 1


        # Do Random Forest
        allScores = np.array(allScores)
        print "SVR Scores" + `c`, allScores
        print("Accuracy: %0.2f (+/- %0.2f)" % (allScores.mean(), allScores.std()))
        print('MeanAbsoluteError Train: {}'.format(metrics.mean_absolute_error(svr_linear.predict(X_train), y_train)))
        print('MeanAbsoluteError Test: {}'.format(metrics.mean_absolute_error(svr_linear.predict(X_test), y_test)))


        allScores = []
        fold = 1
        for train_index, test_index in skf:
            #print("TRAIN:", train_index, "TEST:", test_index)
            X_train, X_test = X[train_index], X[test_index]
            y_train, y_test = y[train_index], y[test_index]

            rndmForest = RandomForestRegressor(n_estimators=100)
            print "Start training Random Forest"
            start_time_training = time.time()
            rndmForest.fit(X_train, y_train)
            print "Done training in ",time.time() - start_time_training, "seconds"
            score = rndmForest.score(X_test, y_test)
            allScores.append(score)
            y_predictTest = rndmForest.predict(X_test)
            y_predictTrain = rndmForest.predict(X_train)

            # #Random Forest
            # res = RandomForestRegressor(n_estimators=100)
            # res = res.fit(X_train, y_train)
            Y_predicted = rndmForest.predict(X_train)
            Y_test_predicted = rndmForest.predict(X_test)


           # print "y_test", y_test
            #print "y_predictTrain", y_predictTrain
            ys = np.vstack([y_train , y_predictTrain]).T
            ys = ys[ys[:, 0].argsort()]
            #print ys
            y_train = ys[:,0]
            y_predictTrain = ys[:,1]

            ys = np.vstack([y_test , y_predictTest]).T
            ys = ys[ys[:, 0].argsort()]
            #print ys
            y_test = ys[:,0]
            y_predictTest = ys[:,1]
            #exit()
            #Images
            plt.close('all')
            plt.figure()
            plt.subplots_adjust(hspace=.5, wspace=.5)
            plt.subplot(1,2,1)
            plt.title('rndmForest train scaled')
            allSampleIndexes=np.arange(len(X_train[:])) #we put the len so to give to each sample a number
            plt.scatter(allSampleIndexes, y_train, c='g', edgecolor='none', s=1, label='train')
            plt.hold('on')
            plt.scatter(allSampleIndexes, y_predictTrain, c='r', edgecolor='none', s=1, label='prediction')
            plt.xlabel('#of rows')
            plt.ylabel('initiation rates (log)')

            plt.subplot(1,2,2)
            allSampleIndexes=np.arange(len(X_test[:])) #we put the len so to give to each sample a number
            plt.scatter(allSampleIndexes, y_test, c='g', edgecolor='none', s=1, label='test')
            plt.hold('on')
            plt.scatter(allSampleIndexes, y_predictTest, c='r', edgecolor='none', s=1, label='prediction')
            plt.xlabel('#of rows')
            plt.ylabel('initiation rates (log)')
            plt.title('rndmForest test')
            plt.legend()
            plt.suptitle('rndmForest_AllFeats'+ `csvFile` +`fold`)
            plt.savefig('correctciandrinidata/rndmForest_AllFeatsFor' + `c` + 'Chrom' + `csvFile` +`fold`+'.svg')
            # plt.show()
            plt.clf()
            fold += 1

        # print "Prediction of X is ", Y_predicted
        allScores = np.array(allScores)
        print "Random Forest Scores" + `c`, allScores
        print("Accuracy: %0.2f (+/- %0.2f)" % (allScores.mean(), allScores.std()))
        print('MeanAbsoluteError Train: {}'.format(metrics.mean_absolute_error(rndmForest.predict(X_train), y_train)))
        print('MeanAbsoluteError Test: {}'.format(metrics.mean_absolute_error(rndmForest.predict(X_test), y_test)))