ShopTRAINING/server/analysis/explanation.py

"""
药店销售预测系统 - 预测解释函数
"""

import numpy as np
from sklearn.linear_model import LinearRegression

def analyze_influencing_factors(product_id, model_type, predictions, features):
    """
    分析影响预测结果的因素
    
    参数:
    product_id: 产品ID
    model_type: 模型类型
    predictions: 预测结果
    features: 特征数据，包含各个特征的值
    
    返回:
    影响因素分析结果字典
    """
    # 如果特征数据为空，则返回空结果
    if features is None or len(features) == 0:
        return {}
    
    # 确保特征数据和预测结果长度匹配
    if len(features) != len(predictions):
        # 如果不匹配，可能是因为预测结果是未来几天的，而特征是历史的
        # 这种情况下，我们只能分析历史数据中的特征与销量的关系
        features = features[-len(predictions):] if len(features) > len(predictions) else features
        predictions = predictions[:len(features)]
    
    # 将特征数据转换为numpy数组
    X = np.array(features)
    y = np.array(predictions)
    
    # 使用线性回归分析特征重要性
    model = LinearRegression()
    model.fit(X, y)
    
    # 获取特征重要性（系数的绝对值）
    importance = np.abs(model.coef_)
    
    # 归一化特征重要性
    total_importance = np.sum(importance)
    if total_importance > 0:
        normalized_importance = importance / total_importance
    else:
        normalized_importance = np.zeros_like(importance)
    
    # 创建特征重要性字典
    feature_names = ['sales', 'price', 'weekday', 'month', 'is_holiday', 'is_weekend', 'is_promotion', 'temperature']
    importance_dict = {feature_names[i]: float(normalized_importance[i]) for i in range(min(len(feature_names), len(normalized_importance)))}
    
    # 按重要性排序
    sorted_importance = sorted(importance_dict.items(), key=lambda x: x[1], reverse=True)
    
    # 找出最重要的特征（重要性大于10%的特征）
    key_factors = [factor for factor, imp in sorted_importance if imp > 0.1]
    
    return {
        'feature_importance': importance_dict,
        'key_factors': key_factors,
        'sorted_importance': sorted_importance
    }

def generate_prediction_explanation(analysis, product_id, model_type):
    """
    根据分析结果生成预测解释文本
    
    参数:
    analysis: 分析结果字典
    product_id: 产品ID
    model_type: 模型类型
    
    返回:
    解释文本
    """
    explanation = f"产品 {product_id} 的 {model_type} 模型预测分析:\n\n"
    
    # 添加趋势分析
    if 'trend' in analysis:
        trend = analysis['trend']
        explanation += f"1. 销量趋势: {trend['trend_type']}\n"
        explanation += f"   - 趋势斜率: {trend['slope']:.4f}\n"
        explanation += f"   - 确定系数 (R²): {trend['r_squared']:.4f}\n"
        explanation += f"   - 波动程度: {trend['volatility_level']} (波动率: {trend['volatility']:.2f})\n\n"
    
    # 添加统计数据
    if 'statistics' in analysis:
        stats = analysis['statistics']
        explanation += "2. 销量统计:\n"
        explanation += f"   - 平均销量: {stats['mean']:.2f}\n"
        explanation += f"   - 中位数销量: {stats['median']:.2f}\n"
        explanation += f"   - 最低销量: {stats['min']:.2f}\n"
        explanation += f"   - 最高销量: {stats['max']:.2f}\n"
        explanation += f"   - 标准差: {stats['std']:.2f}\n\n"
    
    # 添加日环比变化
    if 'day_over_day' in analysis and analysis['day_over_day']:
        dod = analysis['day_over_day']
        avg_change = sum(dod) / len(dod)
        explanation += "3. 日环比变化:\n"
        explanation += f"   - 平均日环比变化: {avg_change:.2f}%\n"
        explanation += f"   - 最大单日增长: {max(dod):.2f}%\n"
        explanation += f"   - 最大单日下降: {min(dod):.2f}%\n\n"
    
    # 添加影响因素分析
    if 'influencing_factors' in analysis:
        factors = analysis['influencing_factors']
        if factors and 'key_factors' in factors and factors['key_factors']:
            explanation += "4. 主要影响因素:\n"
            for i, factor in enumerate(factors['key_factors']):
                importance = factors['feature_importance'][factor]
                explanation += f"   - {factor}: {importance*100:.2f}%\n"
            explanation += "\n"
    
    # 添加总结
    if 'trend' in analysis:
        trend_type = analysis['trend']['trend_type']
        if trend_type == "上升":
            explanation += "总结: 预测显示销量呈上升趋势，"
        elif trend_type == "下降":
            explanation += "总结: 预测显示销量呈下降趋势，"
        else:
            explanation += "总结: 预测显示销量相对平稳，"
        
        if 'influencing_factors' in analysis and analysis['influencing_factors'] and 'key_factors' in analysis['influencing_factors']:
            key_factors = analysis['influencing_factors']['key_factors']
            if key_factors:
                explanation += f"主要受 {', '.join(key_factors)} 因素影响。"
            else:
                explanation += "没有明显的单一影响因素。"
        else:
            explanation += "影响因素分析不可用。"
    
    return explanation