【AI+量化实战 #03】散户焦虑还是看好？用Python给互动易问答做情绪测温

import mairui
import os

api = mairui.Client("<你的证书>")  # 演示证书，请替换成你自己的正式证书
qa = api.stock_interactiveqa("600519", lt=10)   # 最近 10 条
for item in qa:
    print(item["qt"], item["q"][:50])

# 焦虑/负面词
ANXIETY_WORDS = (
    "跌", "亏损", "减持", "退市", "暴雷", "违规", "处罚", "问询",
    "风险", "担忧", "下滑", "停牌", "质押", "爆仓", "诉讼", "索赔",
    "被套", "套牢", "割肉", "补仓", "缩量", "破发", "破净", "商誉",
    "坏账", "计提", "滞涨", "跳水", "闪崩", "操纵", "调查",
)

# 乐观/正面词
OPTIMISM_WORDS = (
    "增长", "回购", "增持", "分红", "利好", "前景", "突破", "创新",
    "布局", "扩张", "合作", "订单", "中标", "预增", "预盈", "送转",
    "龙头", "低估", "机会", "潜力", "新能源", "芯片", "AI",
)


def score_question(q_text: str) -> int:
    """对单条提问打情绪分：+1 乐观，-1 焦虑，0 中性。"""
    if not q_text:
        return 0
    t = q_text.lower()
    pos = sum(1 for w in OPTIMISM_WORDS if w in t)
    neg = sum(1 for w in ANXIETY_WORDS if w in t)
    if pos > neg:
        return 1
    if neg > pos:
        return -1
    return 0

import os
import sys
import time
from datetime import datetime, timedelta
from collections import Counter

import pandas as pd
import mairui

STOCK_POOL = {
    "600519": "贵州茅台", "000858": "五粮液",   "601318": "中国平安",
    "600036": "招商银行", "000333": "美的集团", "002594": "比亚迪",
    "601012": "隆基绿能", "600809": "山西汾酒",
}
QA_LIMIT = 30
LOOKBACK_DAYS = 180
PRICE_WINDOW = 30

ANXIETY_WORDS = (
    "跌", "亏损", "减持", "退市", "暴雷", "违规", "处罚", "问询",
    "风险", "担忧", "下滑", "停牌", "质押", "爆仓", "诉讼", "索赔",
    "被套", "套牢", "割肉", "补仓", "缩量", "破发", "破净", "商誉",
    "坏账", "计提", "滞涨", "跳水", "闪崩", "操纵", "调查",
)
OPTIMISM_WORDS = (
    "增长", "回购", "增持", "分红", "利好", "前景", "突破", "创新",
    "布局", "扩张", "合作", "订单", "中标", "预增", "预盈", "送转",
    "龙头", "低估", "机会", "潜力", "新能源", "芯片", "AI",
)


def score_question(q_text: str) -> int:
    t = q_text.lower()
    pos = sum(1 for w in OPTIMISM_WORDS if w in t)
    neg = sum(1 for w in ANXIETY_WORDS if w in t)
    return 1 if pos > neg else (-1 if neg > pos else 0)


def fetch_interactiveqa(api, codes):
    rows = []
    cutoff = (datetime.now() - timedelta(days=LOOKBACK_DAYS)).strftime("%Y-%m-%d")
    for code, name in codes.items():
        items = api.stock_interactiveqa(code, lt=QA_LIMIT) or []
        for item in items:
            q_text, q_date = item.get("q", ""), item.get("qt", "")
            if q_date >= cutoff:
                rows.append({
                    "code": code, "name": name, "q_date": q_date,
                    "question": q_text, "has_answer": 1 if item.get("a") else 0,
                    "sentiment": score_question(q_text),
                })
        time.sleep(0.15)
    return pd.DataFrame(rows)


def compute_thermometer(qa_df):
    g = qa_df.groupby(["code", "name"])
    result = g.agg(
        total_qa=("question", "count"),
        anxiety=("sentiment", lambda x: (x == -1).sum()),
        optimism=("sentiment", lambda x: (x == 1).sum()),
    ).reset_index()
    result["temperature"] = (
        (result["optimism"] - result["anxiety"]) / result["total_qa"]
    ).round(3)
    result["anxiety_ratio"] = (
        result["anxiety"] / result["total_qa"]
    ).round(3)
    return result


def fetch_recent_returns(api, codes, days=30):
    et = datetime.now().strftime("%Y%m%d")
    st = (datetime.now() - timedelta(days=int(days / 0.7))).strftime("%Y%m%d")
    rows = []
    for code, name in codes.items():
        kline = api.stock_history(code, "d", "n", st=st, et=et) or []
        if len(kline) >= 2:
            df = pd.DataFrame(kline)
            df["c"] = pd.to_numeric(df["c"], errors="coerce")
            df = df.sort_values("t").reset_index(drop=True)
            ret = round((float(df.iloc[-1]["c"]) / float(df.iloc[0]["c"]) - 1) * 100, 2)
            rows.append({"code": code, "name": name, "ret_30d": ret})
        time.sleep(0.1)
    return pd.DataFrame(rows)


api = mairui.Client("<你的证书>")  # 演示证书，请替换成你自己的正式证书
qa_df = fetch_interactiveqa(api, STOCK_POOL)
thermo = compute_thermometer(qa_df)
ret_df = fetch_recent_returns(api, STOCK_POOL, PRICE_WINDOW)

merged = thermo.merge(ret_df, on=["code", "name"], how="left")
merged = merged.sort_values("temperature", ascending=False)
print(merged[["name", "total_qa", "anxiety", "optimism",
              "temperature", "anxiety_ratio", "ret_30d"]].to_string(index=False))