Tool.py 文件源码

def predictNext(self, stock, pred_date_count, train_batch_size=100, use_NN=True):
        trainX, trainY, trainD = self.getRawByCount(pred_date_count-train_batch_size, pred_date_count);
        testX, testY, testD = self.getSingleRaw(pred_date_count)
        testX = testX.reshape(1, -1)
        # print trainX[0]
        # print list(trainX)
        sc = StandardScaler()
        sc.fit(trainX)
        trainX = sc.transform(trainX)
        testX = sc.transform(testX)

        # trainX = np.delete(trainX,0,axis=1)
        # testX = np.delete(testX,0,axis=1)


        fs_method = 'RFC'
        pred_pro=[1,0]

        trainX,testX = featureSelection (trainX, trainY, testX, [], method=fs_method, testmode=False, n_features_to_select=None)

        if use_NN:
            from Power import NNet
            predY = NNet(TrainX=trainX, TrainY=trainY, TestX=testX)
            # print predY
            # pred_pro=[1,0]
        else:
            clf = ExtraTreesClassifier(criterion='gini', n_estimators=150, max_features='auto', n_jobs=4, class_weight='balanced')
            # clf =  DecisionTreeClassifier(class_weight='balanced')
            clf.fit(trainX, trainY)
            predY = clf.predict(testX)
            # pred_pro = (clf.predict_proba(testX) if hasattr(clf, "predict_proba") else clf.decision_function(testX))

        return predY[0], pred_pro[0],testY, testD, 1-clf.score(trainX, trainY)