Estrategia de negociación de fusión inteligente con múltiples indicadores de KNN

normalize(src, length) =>
    mean_val = ta.sma(src, length)
    std_val = ta.stdev(src, length)
    std_val == 0 ? 0.0 : (src - mean_val) / std_val

/*backtest
start: 2025-01-01 00:00:00
end: 2025-09-03 00:00:00
period: 1h
basePeriod: 1h
exchanges: [{"eid":"Futures_Binance","currency":"ETH_USDT","balance":500000}]
args: [["v_input_int_1",5],["v_input_int_2",40],["v_input_int_3",5],["v_input_int_4",5],["v_input_float_1",0.8]]
*/

//@version=6
strategy("Advanced KNN Trading Strategy", overlay=true, max_bars_back=500)

//输入参数配置：设置KNN算法和交易策略的各项参数
k = input.int(5, "KNN邻居数量", minval=3, maxval=50);//K最近邻算法中K值
lookback = input.int(40, "历史数据回望期", minval=50, maxval=500);//历史数据窗口大小
feature_length = input.int(5, "特征计算周期", minval=5, maxval=20);//技术指标计算周期
norm_length = input.int(5, "标准化计算周期", minval=20, maxval=100);//特征标准化周期
prediction_threshold = input.float(0.8, "预测阈值", minval=0.5, maxval=0.9, step=0.1);//交易信号阈值
stop_loss_pct = input.float(2.0, "止损百分比", minval=0.5, maxval=5.0, step=0.1);//止损比例
take_profit_pct = input.float(4.0, "止盈百分比", minval=1.0, maxval=10.0, step=0.1);//止盈比例

//特征工程函数：数据预处理和特征提取
//标准化函数：将原始数据转换为标准正态分布
normalize(src, length) =>
    mean_val = ta.sma(src, length);//计算均值
    std_val = ta.stdev(src, length);//计算标准差
    std_val == 0 ? 0.0 : (src - mean_val) / std_val;//Z-score标准化

//原始技术特征计算：提取价格、成交量、波动率等基础特征
raw_price_momentum = (close - close[feature_length]) / close[feature_length] * 100;//价格动量
raw_rsi = ta.rsi(close, feature_length);//相对强弱指数
raw_volume_ratio = volume / ta.sma(volume, feature_length);//成交量比率
raw_volatility = ta.stdev(ta.change(close), feature_length) / close * 100;//价格波动率

//趋势特征计算：识别价格趋势强度
sma_short = ta.sma(close, 5);//短期均线
sma_long = ta.sma(close, 20);//长期均线
raw_trend_strength = (sma_short - sma_long) / sma_long * 100;//趋势强度

//MACD特征计算：动量指标
[macd_line, signal_line, histogram] = ta.macd(close, 12, 26, 9);//MACD指标
raw_macd_feature = macd_line - signal_line;//MACD差值

//布林带特征计算：价格相对位置
[bb_middle, bb_upper, bb_lower] = ta.bb(close, 20, 2);//布林带
raw_bb_position = (close - bb_lower) / (bb_upper - bb_lower);//价格在布林带中的位置

//特征标准化：将所有特征转换为相同量级
price_momentum = normalize(raw_price_momentum, norm_length);//标准化价格动量
rsi_feature = normalize(raw_rsi, norm_length);//标准化RSI
volume_ratio = normalize(raw_volume_ratio, norm_length);//标准化成交量比率
volatility = normalize(raw_volatility, norm_length);//标准化波动率
trend_strength = normalize(raw_trend_strength, norm_length);//标准化趋势强度
macd_feature = normalize(raw_macd_feature, norm_length);//标准化MACD
bb_position = normalize(raw_bb_position, norm_length);//标准化布林带位置

//目标变量定义：预测标签生成
future_return = (close - close[1]) / close[1] * 100;//未来收益率
target_label = future_return > 0 ? 1.0 : 0.0;//二分类标签：1为上涨，0为下跌

//KNN核心算法：K最近邻预测函数
knn_predict() =>
    var historical_features = array.new<array<float>>();//历史特征向量存储
    var historical_labels = array.new<float>();//历史标签存储
    
    //当前样本特征向量：使用已知的历史数据避免未来函数
    current_features = array.from<float>(
         price_momentum[1], rsi_feature[1], volume_ratio[1], volatility[1], 
         trend_strength[1], macd_feature[1], bb_position[1]
     )
    
    //历史数据收集：维护固定大小的历史样本窗口
    if bar_index >= lookback
        //滑动窗口：移除最旧数据
        if array.size(historical_features) >= lookback
            array.shift(historical_features)
            array.shift(historical_labels)
        
        //添加新的历史样本：使用[2]期特征预测[1]期走势
        hist_features = array.from<float>(
             price_momentum[2], rsi_feature[2], volume_ratio[2], volatility[2],
             trend_strength[2], macd_feature[2], bb_position[2]
         )
        
        array.push(historical_features, hist_features)
        array.push(historical_labels, target_label[1])
    
    //KNN预测计算：基于历史相似样本进行预测
    prediction = 0.0
    if array.size(historical_features) >= k
        distances = array.new<float>();//距离存储数组
        labels = array.new<float>();//对应标签存储数组
        
        //距离计算：计算当前样本与所有历史样本的欧几里得距离
        for i = 0 to array.size(historical_features) - 1
            hist_point = array.get(historical_features, i)
            distance = 0.0
            
            //欧几里得距离计算：各维度差值平方和的平方根
            for j = 0 to array.size(current_features) - 1
                curr_val = array.get(current_features, j)
                hist_val = array.get(hist_point, j)
                distance += math.pow(curr_val - hist_val, 2)
            
            distance := math.sqrt(distance)
            array.push(distances, distance)
            array.push(labels, array.get(historical_labels, i))
        
        //K个最近邻选择：找出距离最小的K个样本
        knn_predictions = array.new<float>();//K个最近邻的标签
        knn_distances = array.new<float>();//K个最近邻的距离
        
        for n = 0 to k - 1
            min_dist = 999999.0
            min_index = 0
            
            //寻找最小距离的样本索引
            for i = 0 to array.size(distances) - 1
                if array.get(distances, i) < min_dist
                    min_dist := array.get(distances, i)
                    min_index := i
            
            //保存K最近邻结果
            array.push(knn_predictions, array.get(labels, min_index))
            array.push(knn_distances, min_dist)
            
            //标记已使用的样本，避免重复选择
            array.set(distances, min_index, 999999.0)
        
        //加权预测：距离越近的样本权重越大
        weighted_sum = 0.0
        weight_total = 0.0
        
        for i = 0 to array.size(knn_predictions) - 1
            distance = array.get(knn_distances, i)
            weight = distance > 0 ? 1.0 / (distance + 0.001) : 1000.0;//反距离权重
            weighted_sum += array.get(knn_predictions, i) * weight
            weight_total += weight
        
        prediction := weight_total > 0 ? weighted_sum / weight_total : 0.5
    
    prediction

//获取KNN预测结果：执行预测算法
knn_prediction = knn_predict()

//交易信号生成：基于预测结果生成买卖信号
long_threshold = prediction_threshold;//多头信号阈值
short_threshold = 1 - prediction_threshold;//空头信号阈值

//交易信号判断：预测概率超过阈值时产生信号
long_signal = knn_prediction > long_threshold;//多头信号：预测上涨概率高
short_signal = knn_prediction < short_threshold;//空头信号：预测下跌概率高

//风险管理：计算止损止盈价格
long_stop_loss = strategy.position_avg_price * (1 - stop_loss_pct / 100);//多头止损价
long_take_profit = strategy.position_avg_price * (1 + take_profit_pct / 100);//多头止盈价
short_stop_loss = strategy.position_avg_price * (1 + stop_loss_pct / 100);//空头止损价
short_take_profit = strategy.position_avg_price * (1 - take_profit_pct / 100);//空头止盈价

//策略执行：开仓和平仓逻辑
//开仓条件：无持仓时根据信号开仓
if long_signal and strategy.position_size == 0
    strategy.entry("Long", strategy.long, comment="KNN多头: " + str.tostring(knn_prediction, "#.##"))

if short_signal and strategy.position_size == 0
    strategy.entry("Short", strategy.short, comment="KNN空头: " + str.tostring(knn_prediction, "#.##"))

//平仓条件：持仓时设置止损止盈
if strategy.position_size > 0
    strategy.exit("Long Exit", "Long", stop=long_stop_loss, limit=long_take_profit)

if strategy.position_size < 0
    strategy.exit("Short Exit", "Short", stop=short_stop_loss, limit=short_take_profit)

//图表可视化：在图表上显示预测结果和信号
plot(knn_prediction, "KNN预测", color=color.blue, linewidth=2);//预测值曲线
hline(prediction_threshold, "多头阈值", color=color.green, linestyle=hline.style_dashed);//多头阈值线
hline(1 - prediction_threshold, "空头阈值", color=color.red, linestyle=hline.style_dashed);//空头阈值线
hline(0.5, "中性线", color=color.gray, linestyle=hline.style_dotted);//中性参考线

//交易信号标记：在图表上标注买卖点
plotshape(long_signal, "多头信号", shape.triangleup, location.belowbar, 
          color=color.green, size=size.small);//多头信号标记
plotshape(short_signal, "空头信号", shape.triangledown, location.abovebar, 
          color=color.red, size=size.small);//空头信号标记

//交易提醒设置：配置交易信号的自动提醒
alertcondition(long_signal, title="KNN多头信号", 
               message="KNN预测多头信号，预测值: {{plot(\"KNN预测\")}}");//多头信号提醒
alertcondition(short_signal, title="KNN空头信号", 
               message="KNN预测空头信号，预测值: {{plot(\"KNN预测\")}}");//空头信号提醒