esmSensor/predictWinLength.py at master · CompNeurosurg/esmSensor · 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
from scipy.stats import spearmanr, pearsonr
from sklearn.linear_model import LinearRegression, Lasso
from sklearn.ensemble import GradientBoostingRegressor, RandomForestRegressor
from sklearn.svm import SVR
from sklearn.model_selection import cross_val_predict, KFold
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
#import seaborn as sns
#sns.set_context('paper')

allSubs = ['110001','110002','110003','110004','110005','110006','110007','110008','110009','110010','110011','110013','110014','110016','110017','110018']
#'110015'
allSubs = ['110018']
phenotype = np.array([2,1,2,2,2,2,1,2,2,1,1,1,2,2,2,1]) # removed 15


outPath = 'C:/data/processed/ESM_pilot/'

esmColumns = ['subjno', 'mood_well', 'mood_down', 'mood_fright', 'mood_tense', 'phy_sleepy', 'phy_tired',
       'mood_cheerf', 'mood_relax', 'thou_concent', 'pat_hallu', 'loc_where',
       'soc_who', 'soc_who02', 'soc_who03', 'act_what', 'act_what02',
       'act_what03', 'act_norpob', 'prob_mobility', 'prob_stillness',
       'prob_speech', 'prob_walking', 'tremor', 'slowness',
       'stiffness', 'tension', 'dyskinesia', 'onoff',
       'medic', 'beep_disturb', '_datetime', '_datetime_e', 'dayno_n', 'beepno_n','duration']
esmInterest = ['mood_well', 'mood_down', 'mood_fright', 'mood_tense', 'phy_sleepy', 'phy_tired',
       'mood_cheerf', 'mood_relax', 'thou_concent', 'pat_hallu', 'act_norpob', 'prob_mobility', 'prob_stillness',
       'prob_speech', 'prob_walking', 'tremor', 'slowness',
       'stiffness', 'tension', 'dyskinesia', 'onoff',
       'medic', 'beep_disturb']

drop=[ 'subjno','soc_who', 'soc_who02', 'soc_who03', 'act_what', 'act_what02', 'loc_where',
       'soc_who', 'soc_who02', 'soc_who03', 'act_what', 'act_what02',
       'act_what03','_datetime', '_datetime_e', 'dayno_n', 'beepno_n','duration','castorID']
esmInterest = ['act_norpob', 'prob_mobility', 'prob_stillness',
       'prob_speech', 'prob_walking','tremor', 'slowness', 'stiffness', 'tension', 'dyskinesia', 'onoff']
esmInterest = ['tremor','onoff']

winLength = [30, 60, 120, 180, 300, 450, 900]
winLength = [900]
folds=10
rs=np.zeros((len(winLength),len(allSubs),len(esmInterest)))
rsFolds = np.zeros((len(winLength),len(allSubs),len(esmInterest),10))
ps=np.zeros((len(winLength),len(allSubs),len(esmInterest)))
rmses=np.zeros((len(winLength),len(allSubs),len(esmInterest)))
rands=np.zeros((len(winLength),len(allSubs),len(esmInterest),2))

for w, wl in enumerate(winLength):
    for s, sub in enumerate(allSubs):

            esmFeatures = pd.read_csv('C:/data/processed/ESM_pilot/' + sub + '_features' + str(wl) + '.csv',index_col=False)
            esmFeatures = esmFeatures.drop(drop, axis=1, errors='ignore')
            #esmFeatures = esmFeatures.dropna(axis=1)
            dat = esmFeatures.drop(esmColumns,axis=1,errors='ignore')
            #dat=esmFeatures.filter(like='BandPower')
            feats = [c for c in dat.keys() if c[-1]!='C' ]
            dat = dat[feats].values
            #feats = [c for c in dat.keys() if c[:9]=='BandPower' ]
            #dat = dat[feats]
            #est=RandomForestRegressor(n_estimators=100)
            est=LinearRegression()
            #est=SVR()
            #est=Lasso(alpha=1.0)
            for eN, e in enumerate(esmInterest):
                x=dat[~np.isnan(esmFeatures[e]),:]
                y=esmFeatures[e][~np.isnan(esmFeatures[e])].values
                #est=LinearRegression()
                #est=RandomForestRegressor(n_estimators=10)
                prediction = np.zeros(y.shape)
                kf = KFold(n_splits=folds)
                for f, (train, test) in enumerate(kf.split(x)):
                        est.fit(x[train,:],y[train])
                        pred = est.predict(x[test,:])
                        prediction[test]=pred
                        rsFolds[w,s,eN, f],_ = pearsonr(pred,y[test])
                rs[w,s,eN], ps[w,s,eN] = pearsonr(prediction,y)
                rmses[w,s,eN] = np.sqrt(np.mean((y-prediction)**2))
                if wl==900:
                    fig, ax = plt.subplots()

                    line1, = plt.plot(np.arange(len(y))[25:], y[25:])
                    line2, = plt.plot(np.arange(len(prediction))[25:],prediction[25:],'--')
                    plt.legend([line1,line2],['Actual','Prediction'])
                    plt.xlabel('Beep')
                    plt.ylabel('ESM Answer')

                    plt.savefig(e + '_prediction.png', dpi=150)
                    plt.show()
                # Establish baseline
                rounds = 1000
                rRands=[]
                rmseRands=[]
                for r in range(rounds):
                    perm = np.random.permutation(y)
                    r, p = pearsonr(perm,y)
                    rRands.append(r)
                    rmse = np.sqrt(np.mean((y-perm)**2))
                    rmseRands.append(rmse)
                alpha=0.05
                rands[w,s,eN,0] = np.sort(rRands)[int(- alpha * rounds)]
                rands[w,s,eN,1] = np.sort(rmseRands)[int(alpha*rounds)]


for eN, e in enumerate(esmInterest):
    fig, ax = plt.subplots()
    line1, = ax.plot(winLength, rs[:,0,eN], 'g-o')
    line2, = ax.plot(winLength, rands[:,0,eN,0], 'r-x')
    line2 = ax.fill_between(winLength, np.zeros((len(winLength))),rands[:,0,eN,0],color='r',alpha=0.5)
    plt.legend([line1, line2],['Prediction', 'Random'])
    plt.xlabel('Window Length (s)')
    plt.ylabel('Pearson r')
    plt.ylim(0,0.5)
    plt.xticks(winLength,[str(w) for w in winLength])
    plt.grid()
    plt.savefig(e + '_corrs.png',dpi=150)
    plt.show()

    fig, ax = plt.subplots()
    line1 = ax.errorbar(winLength, np.mean(rsFolds[:,0,eN,:],axis=1),np.std(rsFolds[:,0,eN,:],axis=1)/np.sqrt(10),color='g', linestyle='-')
    line2, = ax.plot(winLength, rands[:,0,eN,0], 'r-x')
    line2 = ax.fill_between(winLength, np.zeros((len(winLength))),rands[:,0,eN,0],color='r',alpha=0.5)
    plt.legend([line1, line2],['Prediction', 'Random'])
    plt.xlabel('Window Length (s)')
    plt.ylabel('Pearson r')
    plt.ylim(0,0.5)
    plt.xticks(winLength,[str(w) for w in winLength])
    plt.grid()
    #plt.savefig(e + '_corrs.png',dpi=150)
    plt.show()


    fig, ax = plt.subplots()
    line1, = ax.plot(winLength, rmses[:,0,eN],'-o')
    line2, = ax.plot(winLength, rands[:,0,eN,1], '-x')
    plt.legend([line1, line2],['Prediction', 'Random'])
    plt.xlabel('Window Length (s)')
    plt.ylabel('RMSE')
    plt.savefig(e + '_rmse.png',dpi=150)
    plt.show()