TA.TSF() - 时间序列预测

TA.TSF(close, timeperiod=14)

TSF[t] = LINEARREG[t-1] + SLOPE[t-1]

即：当前回归值 + 一个周期的斜率增量

def main():
    while True:
        records = exchange.GetRecords()
        if len(records) < 50:
            Sleep(1000)
            continue
        
        closes = [r['Close'] for r in records]
        
        # 计算预测值
        forecast = TA.TSF(closes, 14)
        
        # 预测值 vs 实际值
        predicted = forecast[-2]  # 上一期的预测
        actual = closes[-1]       # 实际值
        error = ((actual - predicted) / predicted) * 100
        
        Log(f"上期预测: {predicted:.2f}")
        Log(f"实际价格: {actual:.2f}")
        Log(f"预测误差: {error:.2f}%")
        
        # 当前预测下一期
        next_predicted = forecast[-1]
        current_price = closes[-1]
        
        Log(f"当前价格: {current_price:.2f}")
        Log(f"预测下期: {next_predicted:.2f}")
        
        if next_predicted > current_price:
            Log("📈 预测上涨")
        else:
            Log("📉 预测下跌")
        
        Sleep(60000)

def main():
    while True:
        records = exchange.GetRecords()
        if len(records) < 100:
            Sleep(1000)
            continue
        
        closes = [r['Close'] for r in records]
        current_price = closes[-1]
        
        # 预测值和标准差
        forecast = TA.TSF(closes, 20)
        stddev = TA.STDDEV(closes, 20)
        
        # 价格偏离预测值的程度
        deviation = (current_price - forecast[-1]) / stddev[-1]
        
        Log(f"TSF 预测: {forecast[-1]:.2f}")
        Log(f"当前价格: {current_price:.2f}")
        Log(f"偏离: {deviation:.2f}σ")
        
        # 均值回归策略
        if deviation < -2:
            Log("✅ 价格远低于预测（< -2σ），买入")
        elif deviation > 2:
            Log("❌ 价格远高于预测（> +2σ），卖出")
        
        Sleep(60000)