服务好的网站制作,网推是干什么的,中国商标注册网查询网官网,做行业分析的网站对AI炒股感兴趣的小伙伴可加WX#xff1a;caihaihua057200#xff08;备注#xff1a;学校/公司名字方向#xff09;
另外我还有些AI的应用可以一起研究#xff08;我一直开源代码#xff09;
1、引言
在这期内容中#xff0c;我们回到AI预测股票#xff0c;转而探索… 对AI炒股感兴趣的小伙伴可加WXcaihaihua057200备注学校/公司名字方向
另外我还有些AI的应用可以一起研究我一直开源代码
1、引言
在这期内容中我们回到AI预测股票转而探索人工智能技术如何应用于另一个有趣的领域预测A股大盘。
2、AI与股票的关系
在股票预测中AI充当着数据分析和模式识别的角色。虽然无法确保百分之百准确的结果但它为增加预测的洞察力和理解提供了全新的途径。
3、数据收集与处理akshare爬实时上证指数
import akshare as ak
import numpy as np
import pandas as pd
from pandas.tseries.offsets import CustomBusinessDay
from datetime import datetime
import xgboost as xgbdf ak.stock_zh_index_daily_em(symbolsh000001)
数据预处理时间特征转换及时间特征结合K线特征 today datetime.today()
date_str today.strftime(%Y%m%d)
base int(datetime.strptime(date_str, %Y%m%d).timestamp())
change1 lambda x: (int(datetime.strptime(x, %Y%m%d).timestamp()) - base) / 86400
change2 lambda x: (datetime.strptime(str(x), %Y%m%d)).day
change3 lambda x: datetime.strptime(str(x), %Y%m%d).weekday()df[date] df[date].str.replace(-, )
X df[date].apply(lambda x: change1(x)).values.reshape(-1, 1)
X_month_day df[date].apply(lambda x: change2(x)).values.reshape(-1, 1)
X_week_day df[date].apply(lambda x: change3(x)).values.reshape(-1, 1)
XX np.concatenate((X, X_week_day, X_month_day), axis1)[29:]
FT np.array(df.drop(columns[date]))
min_vals np.min(FT, axis0)
max_vals np.max(FT, axis0)
FT (FT - min_vals) / (max_vals - min_vals)window_size 30
num_rows, num_columns FT.shape
new_num_rows num_rows - window_size 1
result1 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.mean(window, axis0)result1[i] window_meanresult2 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.max(window, axis0)result2[i] window_meanresult3 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.min(window, axis0)result3[i] window_meanresult4 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.std(window, axis0)result4[i] window_mean
result_list [result1, result2, result3, result4]
result np.hstack(result_list)XX np.concatenate((XX, result), axis1)
4、预测模型XGboots y1 df[open][29:]
y2 df[close][29:]
y3 df[high][29:]
y4 df[low][29:]
models1 xgb.XGBRegressor()
models2 xgb.XGBRegressor()
models3 xgb.XGBRegressor()
models4 xgb.XGBRegressor()
models1.fit(XX, y1)
models2.fit(XX, y2)
models3.fit(XX, y3)
models4.fit(XX, y4)
5、应用及画图 start_date pd.to_datetime(today)bday_cn CustomBusinessDay(weekmaskMon Tue Wed Thu Fri)
future_dates pd.date_range(startstart_date, periods6, freqbday_cn)
future_dates_str [date.strftime(%Y-%m-%d) for date in future_dates][1:]
future_dates_str pd.Series(future_dates_str).str.replace(-, )
X_x future_dates_str.apply(lambda x: change1(x)).values.reshape(-1, 1)
X_month_day_x future_dates_str.apply(lambda x: change2(x)).values.reshape(-1, 1)
X_week_day_x future_dates_str.apply(lambda x: change3(x)).values.reshape(-1, 1)
XXX np.concatenate((X_x, X_week_day_x, X_month_day_x), axis1)
last_column result[-1:, ]
repeated_last_column np.tile(last_column, (5, 1))
result repeated_last_columnXXX np.concatenate((XXX, result), axis1)
pred1 models1.predict(XXX)
pred2 models2.predict(XXX)
pred3 models3.predict(XXX)
pred4 models4.predict(XXX)y1 np.array(df[open][-30:])
y2 np.array(df[close][-30:])
y3 np.array(df[high][-30:])
y4 np.array(df[low][-30:])
YD np.array(df[date][-30:])data {open: np.concatenate([y1, pred1]),close: np.concatenate([y2, pred2]),high: np.concatenate([y3, pred3]),low: np.concatenate([y4, pred4]),date:np.concatenate([YD,np.array(future_dates_str)])
}df pd.DataFrame(data)import mplfinance as mpf# df[date] pd.date_range(startRQ, periodslen(df))
df[date] pd.to_datetime(df[date])
df.set_index(date, inplaceTrue)
# mpf.plot(df, typecandle, titleStock K-Line)
my_color mpf.make_marketcolors(upred, # 上涨时为红色downgreen, # 下跌时为绿色# edgei, # 隐藏k线边缘# volumein, # 成交量用同样的颜色inheritTrue)my_style mpf.make_mpf_style(# gridaxisboth, # 设置网格# gridstyle-.,# y_on_rightTrue,marketcolorsmy_color)mpf.plot(df, typecandle,stylemy_style,# datetime_format%Y年%m月%d日,titleStock K-Line)
6、结果预测下周上证图中后五天是预测结果 总结图中所示
1、周一到周三略微上涨一点点。
2、下周四五高开高走令人惊讶。
如果提前布局的话应该是选择在周四找最低点买入。
全代码一件运行
import akshare as ak
import numpy as np
import pandas as pd
from pandas.tseries.offsets import CustomBusinessDay
from datetime import datetime
import xgboost as xgbdf ak.stock_zh_index_daily_em(symbolsh000001)today datetime.today()
date_str today.strftime(%Y%m%d)
base int(datetime.strptime(date_str, %Y%m%d).timestamp())
change1 lambda x: (int(datetime.strptime(x, %Y%m%d).timestamp()) - base) / 86400
change2 lambda x: (datetime.strptime(str(x), %Y%m%d)).day
change3 lambda x: datetime.strptime(str(x), %Y%m%d).weekday()df[date] df[date].str.replace(-, )
X df[date].apply(lambda x: change1(x)).values.reshape(-1, 1)
X_month_day df[date].apply(lambda x: change2(x)).values.reshape(-1, 1)
X_week_day df[date].apply(lambda x: change3(x)).values.reshape(-1, 1)
XX np.concatenate((X, X_week_day, X_month_day), axis1)[29:]
FT np.array(df.drop(columns[date]))
min_vals np.min(FT, axis0)
max_vals np.max(FT, axis0)
FT (FT - min_vals) / (max_vals - min_vals)window_size 30
num_rows, num_columns FT.shape
new_num_rows num_rows - window_size 1
result1 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.mean(window, axis0)result1[i] window_meanresult2 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.max(window, axis0)result2[i] window_meanresult3 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.min(window, axis0)result3[i] window_meanresult4 np.empty((new_num_rows, num_columns))
for i in range(new_num_rows):window FT[i: i window_size]window_mean np.std(window, axis0)result4[i] window_mean
result_list [result1, result2, result3, result4]
result np.hstack(result_list)XX np.concatenate((XX, result), axis1)y1 df[open][29:]
y2 df[close][29:]
y3 df[high][29:]
y4 df[low][29:]
models1 xgb.XGBRegressor()
models2 xgb.XGBRegressor()
models3 xgb.XGBRegressor()
models4 xgb.XGBRegressor()
models1.fit(XX, y1)
models2.fit(XX, y2)
models3.fit(XX, y3)
models4.fit(XX, y4)start_date pd.to_datetime(today)bday_cn CustomBusinessDay(weekmaskMon Tue Wed Thu Fri)
future_dates pd.date_range(startstart_date, periods6, freqbday_cn)
future_dates_str [date.strftime(%Y-%m-%d) for date in future_dates][1:]
future_dates_str pd.Series(future_dates_str).str.replace(-, )
X_x future_dates_str.apply(lambda x: change1(x)).values.reshape(-1, 1)
X_month_day_x future_dates_str.apply(lambda x: change2(x)).values.reshape(-1, 1)
X_week_day_x future_dates_str.apply(lambda x: change3(x)).values.reshape(-1, 1)
XXX np.concatenate((X_x, X_week_day_x, X_month_day_x), axis1)
last_column result[-1:, ]
repeated_last_column np.tile(last_column, (5, 1))
result repeated_last_columnXXX np.concatenate((XXX, result), axis1)
pred1 models1.predict(XXX)
pred2 models2.predict(XXX)
pred3 models3.predict(XXX)
pred4 models4.predict(XXX)y1 np.array(df[open][-30:])
y2 np.array(df[close][-30:])
y3 np.array(df[high][-30:])
y4 np.array(df[low][-30:])
YD np.array(df[date][-30:])data {open: np.concatenate([y1, pred1]),close: np.concatenate([y2, pred2]),high: np.concatenate([y3, pred3]),low: np.concatenate([y4, pred4]),date:np.concatenate([YD,np.array(future_dates_str)])
}df pd.DataFrame(data)import mplfinance as mpf# df[date] pd.date_range(startRQ, periodslen(df))
df[date] pd.to_datetime(df[date])
df.set_index(date, inplaceTrue)
# mpf.plot(df, typecandle, titleStock K-Line)
my_color mpf.make_marketcolors(upred, # 上涨时为红色downgreen, # 下跌时为绿色# edgei, # 隐藏k线边缘# volumein, # 成交量用同样的颜色inheritTrue)my_style mpf.make_mpf_style(# gridaxisboth, # 设置网格# gridstyle-.,# y_on_rightTrue,marketcolorsmy_color)mpf.plot(df, typecandle,stylemy_style,# datetime_format%Y年%m月%d日,titleStock K-Line)
