广告效果归因揭秘：为何“过于完美”的数据反而效果差？

/
 * 广告请求参数健康度评估器
 */
public class AdRequestHealthChecker {
    
    /
     * 检查设备ID缺失率是否处于可疑区间
     * @param totalRequests 总请求数
     * @param missingDeviceIdRequests 缺失设备ID的请求数
     * @return HealthStatus 健康状态
     */
    public static HealthStatus checkDeviceIdHealth(int totalRequests, int missingDeviceIdRequests) {
        double missingRate = (double) missingDeviceIdRequests / totalRequests;
        
        // 低缺失率风险区间：缺失率 < 40%
        if (missingRate < 0.4) {
            return HealthStatus.RISKY; // 高风险，可能触发反作弊
        } 
        // 合理缺失区间：40% <= 缺失率 <= 80%
        else if (missingRate <= 0.8) {
            return HealthStatus.HEALTHY; // 健康，属于正常环境波动
        } 
        // 高缺失率区间：缺失率 > 80%
        else {
            return HealthStatus.POOR; // 效果可能差，但因数据不足所致
        }
    }
    
    public enum HealthStatus {
        RISKY,    // 高风险
        HEALTHY,  // 健康
        POOR      // 效果差
    }
}

// 使用示例
public class Main {
    public static void main(String[] args) {
        int totalRequests = 10000;
        int missingRequests = 1500; // 缺失率15%
        
        HealthStatus status = AdRequestHealthChecker.checkDeviceIdHealth(totalRequests, missingRequests);
        if (status == HealthStatus.RISKY) {
            System.out.println("警告：设备ID缺失率过低，存在反作弊风险！");
        }
    }
}

/
 * 用户画像完整性评估器
 */
public class UserProfileIntegrityChecker {
    
    // 定义关键用户画像参数
    private static final List<String> CRITICAL_PROFILE_FIELDS = Arrays.asList(
        "birthTime", "bootTimeSec", "appList", "caids"
    );
    
    /
     * 计算用户画像的完整性评分
     * @param request 广告请求对象
     * @return 完整性得分 (0.0 - 1.0)
     */
    public double calculateProfileScore(AdRequest request) {
        int presentFields = 0;
        
        for (String field : CRITICAL_PROFILE_FIELDS) {
            if (isFieldPresent(request, field)) {
                presentFields++;
            }
        }
        
        return (double) presentFields / CRITICAL_PROFILE_FIELDS.size();
    }
    
    /
     * 判断画像完整性是否达标
     */
    public boolean isProfileSufficient(AdRequest request) {
        double score = calculateProfileScore(request);
        // 如果关键参数缺失超过一半，则认为画像不完整
        return score >= 0.5;
    }
    
    private boolean isFieldPresent(AdRequest request, String field) {
        // 模拟实现：根据字段名检查请求对象中该参数是否存在且有效
        // 实际实现中会通过反射或字段映射来检查
        return request.getFieldValue(field) != null;
    }
}

/
 * 广告位参数健康度综合评分模型
 */
public class AdSlotHealthScorer {
    
    private DeviceIdHealthChecker deviceIdChecker;
    private UserProfileIntegrityChecker profileChecker;
    
    public AdSlotHealthScore calculateHealthScore(AdSlotStats stats) {
        AdSlotHealthScore score = new AdSlotHealthScore();
        
        // 1. 评估设备ID健康度 (权重最高：50%)
        HealthStatus deviceStatus = deviceIdChecker.checkDeviceIdHealth(
            stats.getTotalRequests(), 
            stats.getMissingDeviceIdCount()
        );
        score.setDeviceIdScore(normalizeHealthStatus(deviceStatus));
        
        // 2. 评估用户画像完整性 (权重：30%)
        double profileScore = profileChecker.calculateProfileScore(stats.getSampleRequest());
        score.setProfileScore(profileScore);
        
        // 3. 评估参数采集一致性 (权重：20%)
        double consistencyScore = calculateConsistencyScore(stats);
        score.setConsistencyScore(consistencyScore);
        
        // 计算加权综合分
        double totalScore = (score.getDeviceIdScore() * 0.5) 
                          + (score.getProfileScore() * 0.3)
                          + (score.getConsistencyScore() * 0.2);
        score.setTotalScore(totalScore);
        
        // 设定健康等级
        score.setHealthLevel(determineHealthLevel(totalScore));
        
        return score;
    }
    
    /
     * 计算参数采集一致性得分
     * 原理：检查设备类参数和画像类参数的缺失率是否同步
     */
    private double calculateConsistencyScore(AdSlotStats stats) {
        double deviceMissingRate = (double) stats.getMissingDeviceIdCount() / stats.getTotalRequests();
        double profileMissingRate = (double) stats.getMissingProfileFieldCount() / stats.getTotalRequests();
        
        // 计算两种缺失率的差异，差异越小，一致性越好
        double discrepancy = Math.abs(deviceMissingRate - profileMissingRate);
        
        // 将差异映射到0-1的得分，差异越小得分越高
        return Math.max(0, 1 - discrepancy);
    }
    
    private double normalizeHealthStatus(HealthStatus status) {
        switch (status) {
            case RISKY: return 0.3; // 低缺失率风险，给低分
            case HEALTHY: return 0.9; // 合理缺失，给高分
            case POOR: return 0.5; // 高缺失率，给中等分
            default: return 0.5;
        }
    }
    
    private HealthLevel determineHealthLevel(double totalScore) {
        if (totalScore >= 0.8) return HealthLevel.EXCELLENT;
        else if (totalScore >= 0.6) return HealthLevel.GOOD;
        else if (totalScore >= 0.4) return HealthLevel.WARNING;
        else return HealthLevel.CRITICAL;
    }
    
    public enum HealthLevel {
        EXCELLENT, // 优秀
        GOOD,      // 良好
        WARNING,   // 警告
        CRITICAL   // 严重
    }
}

/
 * 健康度评分结果封装类
 */
class AdSlotHealthScore {
    private double deviceIdScore;      // 设备ID健康度得分
    private double profileScore;       // 用户画像完整性得分
    private double consistencyScore;   // 参数一致性得分
    private double totalScore;         // 综合得分
    private AdSlotHealthScorer.HealthLevel healthLevel; // 健康等级
    
    // ... Getter and Setter methods
}