def up_down_stats(self):
m_df = self.db.load_data(table_name=TableName.DAY,
time_from=self.time_from,
symbols=[self.selected_stock])
m_df_spy = self.db.load_data(table_name=TableName.DAY,
time_from=self.time_from,
symbols=["SPY"])
m_df, m_df_spy = FinI.get_sizes(m_df, m_df_spy)
self.candle_chart(m_df, m_df_spy)
# above_sma9, under_sma9, above_boll,under_boll = {"1":0,"2":0,"3":0,"4":0,"5":0,"6":0}
ind = pd.DataFrame()
ws = pd.DataFrame()
u = 0
pred = pd.DataFrame()
for index, row in m_df.iterrows():
if row.weekday == 0:
i = 1
while m_df.iloc[
m_df.index.get_loc(index) +
i].weekday == m_df.iloc[m_df.index.get_loc(index) + 1 +
i].weekday - 1:
rw = m_df.iloc[m_df.index.get_loc(index) + i]
st.write(str(rw.date) + " - " + str(rw.size_body))
i += 1
st.write(row.week_in_month)
if True or row.size_sma9 > 0 and row.size_boll > 0:
if row.size_body > 0:
# st.write(u)
if u < 0:
ind = ind.append([u])
u = 0
u += 1
elif row.size_body < 0:
if u > 0:
ind = ind.append([u])
# st.write(ind)
u = 0
u -= 1
#
st.dataframe(ind)
st.dataframe(ind.groupby(ind.columns[0]).size())
def prepare_data_for_ml(self):
m_df = self.db.load_data(table_name=TableName.DAY,
time_from=self.time_from,
symbols=[self.selected_stock])
m_df_spy = self.db.load_data(table_name=TableName.DAY,
time_from=self.time_from,
symbols=["SPY"])
m_df, m_df_spy = FinI.get_sizes(m_df, m_df_spy)
st.dataframe(m_df)
self.process_data(m_df)
self.candle_chart(m_df, m_df_spy)
self.show_sizes(m_df)
self.get_heatmap(m_df)
self.nn_pred(m_df)
def logistic_regresion(self, symbols=None):
if symbols is None:
symbols = [self.selected_stock]
# LOGISTIC REGRESSION sucesfully finish this logic
st.write(f"Logistic regression: {self.selected_stock}")
df = self.db.load_data(table_name=TableName.DAY,
time_from=self.time_from,
symbols=symbols)
m_df_spy = self.db.load_data(table_name=TableName.DAY,
time_from=self.time_from,
symbols=["SPY"])
df_best_buy = pd.DataFrame()
df_lr_raw = self.logistic_regression_raw()
st.write("logistic regression RAW SPY")
st.dataframe(df_lr_raw)
# remove volume, industry, symbol cols
# df = df.iloc[:,:4]
# df = df.dropna()
# df = df.iloc[:, :4]
# df = df.retype(spy)
df['open-close'] = df['close'] - df['open'].shift(1)
df['close-close'] = df['close'].shift(-1) - df['close']
# wrong close close only for research
df['close-close-prev'] = df['close'] - df['close'].shift(1)
# df['close-close-1'] = df['close'] - df['close'].shift(1)
# df['close-close-2'] = df['close'].shift(1) - df['close'].shift(2)
# df['close-close-3'] = df['close'].shift(2) - df['close'].shift(3)
# df['close-close-4'] = df['close'].shift(3)- df['close'].shift(4)
df, m_df_spy = FinI.get_sizes(df, m_df_spy)
# df = FinI.add_weekday(df)
# df = FinI.add_week_of_month(df)
df = FinI.add_yearweek(df)
# df = FinI.add_levels(df)
# only first nine rows be aware of this
# df = FinI.add_indicators(df)
# df["MACD"] = ta.trend.macd(close=df['close'])
# df['S_9'] = df['close'].rolling(window=9).mean()
# df['S_20'] = df['close'].rolling(window=20).mean()
# df['S_50'] = df['close'].rolling(window=50).mean()
# df['Corr'] = df['close'].rolling(window=10).corr(df['S_9'])
# df['RSI'] = ta.momentum.rsi(close= df['close'], n=9)
# Initialize Bollinger Bands Indicator
# Add Bollinger Bands features
# df['bb_bbm'] = indicator_bb.bollinger_mavg()
# df['bb_bbh'] = indicator_bb.bollinger_hband()
# df['bb_bbl'] = indicator_bb.bollinger_lband()
df['S_9'] = df['close'].rolling(window=9).mean()
df['S_20'] = df['close'].rolling(window=20).mean()
# df['S_50'] = df['close'].rolling(window=50).mean()
# df['S_200'] = df['close'].rolling(window=200).mean()
df['Corr_9'] = df['close'].rolling(window=9).corr(df['S_9'])
df['Corr_20'] = df['close'].rolling(window=9).corr(df['S_20'])
df['RSI'] = ta.momentum.rsi(close=df['close'])
# df["MACD"] = ta.trend.macd(close=df['close'])
df = df.fillna(0)
st.write("Correlation Matrix - Heatmap")
self.get_heatmap(df)
# remove open, high, low
# df = df.iloc[:,3:]
y = np.where(df['close'].shift(-1) > df['close'], 1, -1)
st.write("Rows:")
st.write(len(df))
# df = df.drop(columns= ["date",
# "yearweek",
# "size_boll_ub",
# "size_boll_lb",
# "size_boll",
# "up_down_row",
# "green_red_row",
# "price_level",
# "weekday",
# "boll",
# "sma9",
# "close_20_sma",
# "size_top",
# "size_btm",
# "close_20_mstd",
# "size_top-2",
# "size_top-1",
# "size_btm-2",
# "size_btm-1",
# "boll_ub",
# "boll_lb",
# "size_btm-3",
# "size_top-3",
# "size_sma9",
# "size_sma20",
# "sma20",
# "open",
# "close",
# "high",
# "low",
# "size_body-1",
# "size_body-2",
# "size_body-3",
# "week_in_month",
# "size_body",
# "size_prev_chng",
# "symbol",
# "sector",
# "sym",
# "industry",
# "amount",
# "volume",
# "open-close",
# "close-close"
# ])
df_strategy = df.copy()
# relatively good negative prediction
# df = df[["size_sma9"]]
# values by corelation matrix
# df = df[["size_sma20","size_sma9","up_down_row"]]
# df = df[["green_red_row", "size_sma9", "up_down_row", "size_body"]]
# df = df[["green_red_row", "size_sma9", "up_down_row", "size_body","volume"]]
# origin
# df = df[["Corr_9","Corr_20", "open-close", "close-close-prev", "RSI","sma9"]]
df = df[["Corr_9", "open-close", "close-close-prev", "RSI", "sma9"]]
# overnight, negative precision
# df = df[["green_red_row", "size_body-1","size_body-2","size_body-3", "size_prev_chng","weekday"]]
# df = df[["close", "sma9", "sma20"]]
st.write("Dataframe")
st.dataframe(df)
X = df.iloc[:, :30]
st.write(len(y))
st.write(len(X))
# split to test dataset
split = int(0.7 * len(df))
X_train, X_test, y_train, y_test = X[:split], X[split:], y[:split], y[
split:]
# We will instantiate the logistic regression in Python using ‘LogisticRegression’
# function and fit the model on the training dataset using ‘fit’ function.
model = LogisticRegression()
model = model.fit(X_train, y_train)
# Examine coeficients
pd.DataFrame(zip(X.columns, np.transpose(model.coef_)))
st.write("Examine The Coefficients")
st.write(pd.DataFrame(zip(X.columns, np.transpose(model.coef_))))
#We will calculate the probabilities of the class for the test dataset using ‘predict_proba’ function.
st.write("probability")
probability = model.predict_proba(X_test)
st.write(probability)
st.write("predicted")
predicted = model.predict(X_test)
st.write(predicted)
st.write("Y test")
st.write(y_test)
st.write(len(X_test))
st.write("Confusion matrix")
conf_matrix = metrics.confusion_matrix(y_test, predicted)
st.write(conf_matrix)
st.write("classification report")
st.write(metrics.classification_report(y_test, predicted))
st.write("model accuracy")
st.write(model.score(X_test, y_test))
st.write("cross validation accuracy")
cross_val = cross_val_score(LogisticRegression(),
X,
y,
scoring='accuracy',
cv=10)
st.write(cross_val)
st.write(cross_val.mean())
st.write("Trading Strategy")
df['Predicted_Signal'] = model.predict(X)
st.write("Predicted signals")
st.dataframe(df)
#STRARTEGY count
st.dataframe(df_strategy)
# df_strategy = df_strategy.tail(len(df_strategy) - split)
df_strategy["Predicted_Signal"] = model.predict(X)
price_change = df_strategy.iloc[len(df_strategy) -
1].close - df_strategy.iloc[0].close
st.write("Perfect gain")
df_strategy["perfect_gain"] = df_strategy["close-close"].where(
df_strategy["close-close"] > 0).sum()
st.write(df_strategy["perfect_gain"].iloc[0])
st.write("Invest Gain")
st.write(price_change)
st.write("L.G. gain")
df_strategy["lg_gain"] = df_strategy["close-close"].where(
df_strategy["Predicted_Signal"] > 0).sum()
st.write(df_strategy["lg_gain"].iloc[0])
gap = -2
st.write("sma9 gain: " + str(gap))
df_strategy["sma9_gain"] = df_strategy["close-close"].where(
df_strategy["size_sma9"] > gap).sum()
st.write(df_strategy["sma9_gain"].iloc[0])
st.write("sma9 gain 0")
df_strategy["sma9_gain"] = df_strategy["close-close"].where(
df_strategy["size_sma9"] > 0).sum()
st.write(df_strategy["sma9_gain"].iloc[0])
st.write("sma20 gain")
df_strategy["sma20_gain"] = df_strategy["close-close"].where(
df_strategy["size_sma20"] > -2).sum()
st.write(df_strategy["sma20_gain"].iloc[0])