claraleeee/hftpredict

1# app.py2import streamlit as st3from predict import predict_for, build_a_stock_minutes, build_us_minutes, trading_minutes_for_date4import matplotlib.pyplot as plt5import pandas as pd6 7st.set_page_config(page_title="📈 股票预测系统", layout="wide")8st.title("📈 股票价格预测系统")9 10market = st.radio("选择市场类型", ["A股 / ETF", "美股"])11code = st.text_input("输入股票代码", "sh601988" if market == "A股 / ETF" else "AAPL")12date = st.text_input("预测日期 (YYYY-MM-DD)", "2025-11-13")13open_price = st.number_input("开盘价", value=5.77 if market == "A股 / ETF" else 510.0)14run = st.button("运行预测")15 16if run:17    mkt_key = "a" if market == "A股 / ETF" else "us"18    outs = predict_for(mkt_key, code, date, open_price=open_price)19 20    # build prediction span bounds from predicted timelines only21    pred_mins = [pd.to_datetime(df["datetime"].min()) for df in outs]22    pred_maxs = [pd.to_datetime(df["datetime"].max()) for df in outs]23    pred_min = min(pred_mins)24    pred_max = max(pred_maxs)25 26    # fetch actual data but then restrict to prediction window27    if market == "A股 / ETF":28        df_actual = build_a_stock_minutes(code, days=3)  # fetch a few days to ensure coverage29    else:30        df_actual = build_us_minutes(code, days=3)31    # filter actual to the pred window (so old 15:00 won't affect x axis)32    df_actual = df_actual[(pd.to_datetime(df_actual["datetime"]) >= pred_min) & 33                          (pd.to_datetime(df_actual["datetime"]) <= pred_max)].reset_index(drop=True)34 35    # plotting36    fig, ax = plt.subplots(figsize=(12, 5))37    if not df_actual.empty:38        ax.plot(pd.to_datetime(df_actual["datetime"]), df_actual["close"], label="实际价格", color="gray", linewidth=1.5)39 40    ymin = None; ymax = None41    for df in outs:42        times = pd.to_datetime(df["datetime"])43        prices = pd.to_numeric(df["predicted_price"], errors="coerce")44        ax.plot(times, prices, label=df["model"], linewidth=2)45        if ymin is None or (prices.min() < ymin):46            ymin = float(prices.min())47        if ymax is None or (prices.max() > ymax):48            ymax = float(prices.max())49 50    # if actual present, include its range as well51    if not df_actual.empty:52        ymin = min(ymin if ymin is not None else float(df_actual["close"].min()), float(df_actual["close"].min()))53        ymax = max(ymax if ymax is not None else float(df_actual["close"].max()), float(df_actual["close"].max()))54 55    # set xlim to prediction window explicitly to avoid previous-day bleed56    ax.set_xlim(pred_min, pred_max)57 58    if ymin is not None and ymax is not None:59        margin = max((ymax - ymin) * 0.03, 0.01)60        ax.set_ylim(ymin - margin, ymax + margin)61 62    ax.legend(loc="upper left", bbox_to_anchor=(1.01, 1), borderaxespad=0)63    ax.set_title(f"{market} {code} 预测日期：{date}")64    ax.set_xlabel("时间")65    ax.set_ylabel("价格")66    st.pyplot(fig)67