modify

stock/compont_predict.py

@@ -1,5 +1,39 @@
 from stock import cnn_predict
 from stock import dnn_predict
+from stock import kmeans
+
+
+def dnn_and_kmeans():
+    # cnn_result = cnn_predict.predict()
+    dnn_result = dnn_predict.predict(file_path='D:\\data\\quantization\\stock6_5_test.log', model_path='5d_dnn_seq.h5')
+    cnn_result = kmeans.class_fic(file_path="D:\\data\\quantization\\stock2_20.log")
+    print('计算完成')
+
+    with open('dnn_and_kmeans.txt', 'a') as f:
+        up_num = 0
+        up_right = 0
+        i = 0
+        for m in cnn_result: #3位 时间是字符串
+            # print('find', m)
+            for n in dnn_result: #5位 时间是int
+                if m[0][0] == n[0][0] and int(m[0][1]) == n[0][1]:
+                    print('AND', m)
+                    f.write(str(m) + '\n')
+
+                    if n[1][0] == 1:
+                        up_right = up_right + 1.13
+                    elif n[1][1] == 1:
+                        up_right = up_right + 1.05
+                    elif n[1][2] == 1:
+                        up_right = up_right + 1
+                    else:
+                        up_right = up_right - 0.15
+                    up_num = up_num + 1
+                    i = i + 1
+
+                    break
+
+    print(up_right, up_num, up_right / up_num)
 
 
 def and_predict():
@@ -30,6 +64,6 @@ def and_predict():
 
     print(up_right, up_num, up_right / up_num)
 
-
 if __name__ == '__main__':
-    and_predict()
+    # and_predict()
+    dnn_and_kmeans()

stock/dnn_predict.py

@@ -1,9 +1,5 @@
-import keras
 # -*- encoding:utf-8 -*-
 import numpy as np
-from keras.models import Sequential
-from keras.layers import Dense,Dropout
-import random
 from keras.models import load_model
 
 
@@ -38,9 +34,25 @@ def predict(file_path='', model_path='15min_dnn_seq.h5'):
                 f.write(str([lines[i][-2], lines[i][-1]]) + "\n")
                 win_dnn.append([lines[i][-2], lines[i][-1]])
                 if fact[0] == 1:
+                    up_right = up_right + 1.15
+                elif fact[1] == 1:
+                    up_right = up_right + 1.05
+                elif fact[2] == 1:
                     up_right = up_right + 1
+                else:
+                    up_right = up_right - 0.15
+                up_num = up_num + 1
+            elif r[1] > 0.5:
+                f.write(str([lines[i][-2], lines[i][-1]]) + "\n")
+                win_dnn.append([lines[i][-2], lines[i][-1]])
+                if fact[0] == 1:
+                    up_right = up_right + 1.15
                 elif fact[1] == 1:
-                    up_right = up_right + 0.2
+                    up_right = up_right + 1.05
+                elif fact[2] == 1:
+                    up_right = up_right + 1
+                else:
+                    up_right = up_right - 0.15
                 up_num = up_num + 1
 
             i = i + 1

stock/dnn_train.py

@@ -12,7 +12,7 @@ from imblearn.over_sampling import RandomOverSampler
 def read_data(path):
     lines = []
     with open(path) as f:
-        for x in range(60000):
+        for x in range(30000):
             lines.append(eval(f.readline().strip()))
 
     random.shuffle(lines)
@@ -41,19 +41,20 @@ def train(input_dim=400, result_class=3, file_path="D:\\data\\quantization\\stoc
 
     model = Sequential()
     model.add(Dense(units=120+input_dim, input_dim=input_dim,  activation='relu'))
-    # model.add(Dense(units=60+int(input_dim/2), activation='relu'))
     model.add(Dense(units=120+input_dim, activation='relu',kernel_regularizer=regularizers.l2(0.001)))
+    model.add(Dense(units=120+input_dim, activation='relu'))
     model.add(Dropout(0.2))
-    model.add(Dense(units=60+input_dim, activation='relu'))
+    model.add(Dense(units=60+input_dim, activation='selu'))
     model.add(Dropout(0.2))
     model.add(Dense(units=60+input_dim, activation='selu'))
     # model.add(Dropout(0.2))
-    model.add(Dense(units=512, activation='selu'))
+    model.add(Dense(units=512, activation='relu'))
+
     model.add(Dense(units=result_class, activation='softmax'))
     model.compile(loss='categorical_crossentropy', optimizer="adam",metrics=['accuracy'])
 
     print("Starting training ")
-    h=model.fit(train_x, train_y, batch_size=32, epochs=126, shuffle=True)
+    model.fit(train_x, train_y, batch_size=32, epochs=56, shuffle=True)
     score = model.evaluate(test_x, test_y)
     print(score)
     print('Test score:', score[0])
@@ -69,5 +70,5 @@ def train(input_dim=400, result_class=3, file_path="D:\\data\\quantization\\stoc
 
 
 if __name__ == '__main__':
-    train(input_dim=46, result_class=5, file_path="D:\\data\\quantization\\stock6_5.log", model_name='5d_dnn_seq.h5')
+    train(input_dim=86, result_class=5, file_path="D:\\data\\quantization\\stock6_5.log", model_name='5d_dnn_seq.h5')
     # train(input_dim=400, result_class=3, file_path="D:\\data\\quantization\\stock6.log", model_name='15m_dnn_seq.h5')

stock/kmeans.py

@@ -0,0 +1,96 @@
+# -*- encoding:utf-8 -*-
+from sklearn.cluster import KMeans
+import numpy as np
+from annoy import AnnoyIndex
+
+
+def read_data(path):
+    lines = []
+    with open(path) as f:
+        for x in range(160000):
+            line = eval(f.readline().strip())
+            # if line[-1][0] == 1 or line[-1][1] == 1:
+            lines.append(line)
+
+    return lines
+
+length = 20
+def class_fic(file_path=''):
+    lines = read_data(file_path)
+    print('读取数据完毕')
+    size = len(lines[0])
+    train_x = np.array([s[:length] for s in lines])
+    train_y = [s[size - 1] for s in lines]
+    v_x = train_x.reshape(train_x.shape[0], 4*length)
+    stock_list = [s[size - 2] for s in lines]
+
+    # annoy_sim(v_x)
+    print('save数据完毕')
+    return find_annoy(train_y, stock_list)
+
+def annoy_sim(lines):
+    tree = 30
+    t = AnnoyIndex(length*4, metric="angular") # 24是向量维度
+
+    i = 0
+    for stock in lines:
+        t.add_item(i, stock)
+        i = i + 1
+
+    t.build(tree)
+    t.save('stock_20d.ann')
+
+
+def find_annoy(lines, stock_list):
+    t = AnnoyIndex(length*4, metric="angular")
+    t.load('stock_20d.ann')
+    num = 0
+    right = 0
+
+    win_dnn = []
+    for i in range(len(lines)):
+        index, distance = t.get_nns_by_item(i, 10, include_distances=True)
+        # print(index, distance)
+
+        # 预测
+        total = 0
+        g = 0
+        for j in range(1, len(index)):
+            if distance[j] < 0.4:
+                total = total + 1
+                if lines[j][0] == 1:
+                    g = g + 1
+                elif lines[j][1] == 1:
+                    g = g + 1
+                elif lines[j][2] == 1:
+                    g = g + 0.5
+        if total > 1 and g / total > 0.38:
+            right = right + 1
+            if stock_list[i][1] > 20181101:
+                print(stock_list[i])
+                win_dnn.append([stock_list[i], lines[i]])
+
+
+        # 计算
+        # if lines[i][0] == 1:
+        #     g = 0
+        #     total = 0
+        #     for j in range(1,len(index)):
+        #         if distance[j] < 0.4:
+        #             total = total + 1
+        #             if lines[j][0] == 1:
+        #                 g = g+1
+        #             elif lines[j][1] == 1:
+        #                 g = g+1
+        #     if total > 1 and g/total > 0.21:
+        #         right = right + 1
+        #     if total > 1:
+        #         num = num + 1
+
+    print(right, num)
+    print('find数据完毕')
+    return win_dnn
+
+
+if __name__ == '__main__':
+    class_fic(file_path="D:\\data\\quantization\\stock2_20.log")