Quant-Models/tsmom.py at master · turing-usp/Quant-Models · 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
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

from sklearn.linear_model import LinearRegression
from sklearn.tree         import DecisionTreeRegressor
from sklearn.ensemble     import RandomForestRegressor

from turing_quant_models import Turing_quant_models


class tsmom (Turing_quant_models):

    def __init__(self, df):

        Turing_quant_models.__init__(self, df)

    def ml_signal(self, model, prices, lags=[3, 6, 12]):
        """
        Return predict and signal construction (long or short)
        """
        signal = []

        for i in prices.columns:
            # converting Series to DataFrame to do the feature engineering
            stock = pd.DataFrame(prices[i].copy())

            # renaming target variable
            stock.columns = ["y"]

            X = []
            y = []

            # adding lagged columns
            for i in lags:
                stock["lag_{}".format(i)] = stock.y.pct_change(i).fillna(0)

            y = stock[-24:-2].dropna().y                   # target
            X = stock[-24:-2].dropna().drop(['y'], axis=1) # features

            try:

                model.fit(X, y)

                X_pred = stock.iloc[-1,1:]

                y_pred = model.predict([X_pred])

                sign = np.where(y_pred/stock.y.iloc[-1] - 1 > 0, 1, -1)

                signal.append(sign[0])

            except:

                signal.append(0)


            print(signal)

        return signal

    def signal(self, df, date, passive, method="momentum"):
        """
        Função que constrói o sinal para diversos metodos, sendo o tradicional: momentum
        """

        num_assets = len(df.iloc[-1])
        signal = []

        if passive:
            signal = np.ones(num_assets)
        else:

            if method == "momentum":

                returns = df.pct_change(21 * 12).resample('BM').last().ffill()[:date]

                signal = np.where(returns.iloc[-1] > 0, 1, -1)

            elif method == "momentum_lagged":

                returns_12 = df.pct_change(21 * 12).resample('BM').last().ffill()[:date]

                returns_6 = df.pct_change(21 * 6).resample('BM').last().ffill()[:date]

                returns_3 = df.pct_change(21 * 3).resample('BM').last().ffill()[:date]

                momentum_mean = (returns_12.iloc[-1] + returns_6.iloc[-1] + returns_3.iloc[-1]) / 3

                signal = np.where(momentum_mean > 0, 1, -1)

            elif method == "linear_regression":

                prices = df.resample('BM').last().ffill()[:date]

                lr = LinearRegression()

                signal = self.ml_signal(lr, prices)

            elif method == "decision_tree":

                prices = df.resample('BM').last().ffill()[:date]

                dt = DecisionTreeRegressor()

                signal = self.ml_signal(dt, prices)

            elif method == "random_forest":

                prices = df.resample('BM').last().ffill()[:date]

                rf = RandomForestRegressor()

                signal = self.ml_signal(rf, prices)

        return signal

    def tsmom(self, df, returns_monthly, vol_monthly, date, method='momentum', risk=0.4, passive=False, momentum_window=12):

        position = self.signal(df, date, passive, method)

        weights = (risk / vol_monthly.iloc[date-1])

        weights /= len(weights)

        portfolio = position * weights

        return (1+np.dot(portfolio, returns_monthly.iloc[date]))

    def backtesting(self, start_date, years, vol, method, plot=True):

        returns_model = []  # retorno do TSMOM
        returns_baseline = []  # retorno passivo
        start = start_date
        years = years
        end = 12*(int(start/12) + years)

        for i in range(start, end):

            self.printProgressBar (i-start, end-start-1)

            returns_model.append(self.tsmom(self.close_df, self.returns_monthly,
                                            self.vol_monthly, i))

            returns_baseline.append(self.tsmom(self.close_df, self.returns_monthly,
                                               self.vol_monthly, i, passive=True))

        returns_model = pd.DataFrame(returns_model)
        returns_baseline = pd.DataFrame(returns_baseline)

        returns_model.index = self.returns_monthly.iloc[start:end].index
        returns_baseline.index = self.returns_monthly.iloc[start:end].index

        if plot:
            self.plot_backtesting(
                returns_model, returns_baseline, "TSMOM", "Long only", "Cumulative returns")