stockbot_2020/smabc.py at master · billray0259/stockbot_2020 · 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
import pandas as pd
import numpy as np
import pickle
from tqdm import tqdm
import os
import math

from data_handler import DataHandler

import matplotlib.pyplot as plt

class SimpleMovingAverageBinaryClassifier:

    def __init__(self, name, sub_dir=None):
        self.data_handler = DataHandler(name)
        self.dir = os.path.join(self.data_handler.data_dir, "smabc_training")
        if sub_dir is not None:
            self.dir = os.path.join(self.dir, sub_dir)
        self.model_name = "model.hdf5"
        self.model_file = os.path.join(self.dir, self.model_name)
        print(self.model_file)

        self.model = None
        if os.path.exists(self.model_file):
            import keras
            self.model = keras.models.load_model(self.model_file)

    def preprocess(self, candles=None, pred_range=208, density=0.8, win1_start=1, win1_end=256, win1_step=8, win2_start=2, win2_end=1024, win2_step=32):
        live = candles is not None
        candles = pd.read_hdf(self.data_handler.filled_file) if not live else candles
        opens = candles.filter(like="_open")
        closes = candles.filter(like="_close")

        opens.columns = list(map(lambda column: column.split("_")[0], opens.columns))
        closes.columns = list(map(lambda column: column.split("_")[0], closes.columns))
        returns = np.log(opens/closes)
        returns.fillna(value=0, inplace=True)  # pylint: disable=no-member

        vectors = []
        labels = []
        if live:
            iterator = range(len(closes)-1, len(closes))
        else:
            iterator = tqdm(range(win2_end, len(closes)-pred_range, 10))
        for i in iterator:
            features = []

            if not live:
                labels.append(returns[i:i+pred_range].sum() > 0)

            win1 = win1_start
            while win1 < win1_end:
                win2 = win2_start
                while win2 < win2_end:
                    if win1 < win2:
                        sma1 = closes[i-win1:i].mean()
                        sma2 = closes[i-win2:i].mean()
                        vec = np.where(sma1 > sma2, 1, -1)
                        features.append(vec)

                    win2 += max(round((1-density) * win2), win1_step)
                win1 += max(round((1-density) * win1), win2_step)

            features = np.array(features)
            vectors.append(np.transpose(features))

        vectors = np.array(vectors)
        x = np.concatenate(vectors)
        if live:
            return x

        for i, label in enumerate(labels):
            labels[i] = list(map(lambda x: [0, 1] if x else [1, 0], label))

        labels = np.array(labels)
        y = np.concatenate(labels)

        if not os.path.exists(self.dir):
            os.makedirs(self.dir)

        x_file = os.path.join(self.dir, "x.npy")
        y_file = os.path.join(self.dir, "y.npy")
        np.save(x_file, x)
        np.save(y_file, y)

    def train(self, epochs=25):
        from keras.callbacks import ModelCheckpoint
        from keras.layers import Dense, Dropout
        from keras.models import Sequential
        from keras.activations import softmax
        import os
        os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'
        x_file = os.path.join(self.dir, "x.npy")
        y_file = os.path.join(self.dir, "y.npy")
        x = np.load(x_file)
        y = np.load(y_file)

        num_batches, num_features = x.shape

        model = Sequential([
            Dense(8, activation="relu", input_shape=[num_features]),
            Dropout(0.5),
            Dense(4, activation="relu"),
            Dropout(0.5),
            Dense(2, activation="softmax")
        ])

        model.compile(loss="binary_crossentropy", optimizer="adam", metrics=["accuracy"])
        model.summary()

        filepath = os.path.join(self.dir, "model_{val_accuracy:.4f}.hdf5")
        # filepath = os.path.join(self.dir, "model.hdf5")
        checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
        callbacks = [checkpoint]

        model.fit(x, y, epochs=epochs, batch_size=round(math.sqrt(num_batches)), validation_split=0.2, callbacks=callbacks)

    def wanted_tickers(self):
        return ["TSLA"]

    def get_holdings(self, data, current_holdings):
        x = self.preprocess(data)
        model = self.model
        pred = model.predict(x)[0]
        return {"TSLA": np.argmax(pred)}


if __name__ == "__main__":
    for i in tqdm(range(1, 27, 1)):
        smabc = SimpleMovingAverageBinaryClassifier("stocks_only", sub_dir=str(i))
        smabc.preprocess(pred_range=i)
        smabc.train()