jyps/project_chuan/score.go

package main

import (
	"fmt"
	"math"
	"strings"
	"time"

	"github.com/adrg/strutil"
	"github.com/adrg/strutil/metrics"
)

type ScoredResult struct {
	Document   EsDocument
	Confidence float64
}

type InputData struct {
	ProjectCode       string
	Area              string
	City              string
	ProjectName       string
	ConstructionUnit  string
	AreaCode          string
	PlannedInvestment float64
	BaseDate          time.Time
}

type EsDocument struct {
	Id          string  `json:"id"`
	Title       string  `json:"title"`
	Projectname string  `json:"projectname"`
	ProjectCode string  `json:"projectcode"`
	Toptype     string  `json:"toptype"`
	Subtype     string  `json:"subtype"`
	Area        string  `json:"area"`
	City        string  `json:"city"`
	Buyer       string  `json:"buyer"`
	Budget      float64 `json:"budget"`
	Bidamount   float64 `json:"bidamount"`
	Winner      string  `json:"winner"`
	Detail      string  `json:"detail"`
	Publishtime int64   `json:"publishtime"`
	Agency      string  `json:"agency"`
	SWinner     string  `json:"s_winner"`
	WinnerTel   string  `json:"winnertel"`
	BuyerTel    string  `json:"buyertel"`
	BuyerPerson string  `json:"buyerperson"`
	Score       float64 `json:"score"`
}

// calculateConfidenceScore calculates a score from 0-100 based on multiple factors.
func calculateConfidenceScore(target InputData, candidate EsDocument) float64 {
	var totalScore float64 = 0.0
	nameSimilarity := EnhancedSimilarity(strings.ReplaceAll(target.ProjectName, target.ConstructionUnit, ""), strings.ReplaceAll(strings.ReplaceAll(candidate.Projectname, candidate.Buyer, ""), target.ConstructionUnit, ""))
	// 1. Project Name Similarity (Weight: 40 points)
	totalScore += nameSimilarity * 70.0
	//log.Println(target.ProjectName, candidate.Projectname, nameSimilarity)

	// 2. Construction Unit Similarity (Weight: 30 points)
	unitSimilarity := EnhancedSimilarity(target.ConstructionUnit, candidate.Buyer)
	totalScore += unitSimilarity * 10.0

	//log.Println(target.ConstructionUnit, candidate.Buyer, unitSimilarity)

	if target.Area != "" && target.Area == candidate.Area {
		totalScore += 5
		if target.City != "" && target.City == candidate.City {
			totalScore += 5
		}
	}

	// 3. Project Code Presence
	if target.ProjectCode != "" && strings.Contains(candidate.Detail, target.ProjectCode) {
		totalScore += 5.0
	}

	// 4. 投资额
	if target.PlannedInvestment > 0 && candidate.Budget > 0 {
		diff := math.Abs(target.PlannedInvestment - candidate.Budget)
		percentDiff := diff / math.Max(target.PlannedInvestment, 1) // Avoid division by zero
		// Simple linear decay: 100% score for 0% diff, 0% score for 100% diff or more
		investmentScore := math.Max(0, (1.0-percentDiff)*5.0)
		totalScore += investmentScore
	}

	//计算时间
	if target.BaseDate.Year() > 2000 && candidate.Publishtime > 1600000000 {
		totalScore += TimeProximity(target.BaseDate, time.Unix(candidate.Publishtime, 0)) * 5
	}

	return totalScore
}

// 计算时间
func TimeProximity(t1, t2 time.Time) float64 {
	// 解析时间
	// 计算天数差值(取绝对值)
	days := int(t1.Sub(t2).Abs().Hours() / 24)
	// 根据天数差值返回不同的相似度
	switch {
	case days < 100:
		return 1.0
	case days < 200:
		return 0.7
	case days < 365:
		return 0.55
	case days < 700:
		return 0.3
	default:
		return 0.15
	}
}

func EnhancedSimilarity(str1, str2 string) float64 {
	// 标准化：去除空格和标点（根据中文需求调整）
	s1 := strings.TrimSpace(str1)
	s2 := strings.TrimSpace(str2)

	// 1. 检查完全相等
	if s1 == "" || s2 == "" {
		return 0
	} else if s1 == s2 {
		return 1.0
	}

	// 2. 检查包含关系（权重50%）
	if containmentScore := checkContainment(s1, s2); containmentScore > 0 {
		return containmentScore
	}

	// 3. 使用改进的Jaro-Winkler（考虑中文分词）
	jw := metrics.NewJaroWinkler()
	baseScore := strutil.Similarity(s1, s2, jw)

	// 4. 添加公共子串权重
	substringScore := checkLongestCommonSubstring(s1, s2)

	// 5. 组合得分
	finalScore := 0.5*baseScore + 0.5*substringScore

	return finalScore
}

// 检查包含关系，返回0-1的得分
func checkContainment(s1, s2 string) float64 {
	// 确保s1是较短的字符串
	if len(s1) > len(s2) {
		s1, s2 = s2, s1
	}

	// 完全包含
	if strings.Contains(s2, s1) {
		// 被包含部分占长字符串的比例
		ratio := float64(len(s1)) / float64(len(s2))
		// 基础分0.7 + 比例调整(最多加到0.95)
		return 0.9 + ratio*0.1
	}
	return 0
}

// 检查最长公共子串
func checkLongestCommonSubstring(s1, s2 string) float64 {
	m := len(s1)
	n := len(s2)

	// 动态规划表
	dp := make([][]int, m+1)
	for i := range dp {
		dp[i] = make([]int, n+1)
	}

	maxLen := 0
	for i := 1; i <= m; i++ {
		for j := 1; j <= n; j++ {
			if s1[i-1] == s2[j-1] {
				dp[i][j] = dp[i-1][j-1] + 1
				if dp[i][j] > maxLen {
					maxLen = dp[i][j]
				}
			}
		}
	}

	// 计算得分：最长公共子串占两个字符串平均长度的比例
	avgLen := (m + n) / 2
	if avgLen == 0 {
		return 0
	}
	return float64(maxLen) / float64(avgLen)
}

func main000() {
	str1 := "绿色低碳节能项目"
	str2 := "河北纵横集团丰南钢铁有限公司绿色低碳节能项目"

	similarity := EnhancedSimilarity(str1, str2)
	fmt.Printf("相似度: %.4f\n", similarity)

	// 更多测试用例
	testCases := []struct {
		s1, s2 string
	}{
		{"绿色低碳", "绿色低碳节能项目"},
		{"钢铁项目", "丰南钢铁有限公司项目"},
		{"节能环保", "新能源开发"},
		{"相同的字符串", "相同的字符串"},
	}

	for _, tc := range testCases {
		sim := EnhancedSimilarity(tc.s1, tc.s2)
		fmt.Printf("'%s' vs '%s': %.4f\n", tc.s1, tc.s2, sim)
	}
}