Почему объяснение оценки не лучший результат?

Question

Я делаю программу планирования с использованием optaplanner 7.12, но обнаружил, что иногда объяснение оценки не совпадает с лучшим оценкой, ниже приведен пример, пожалуйста, помогите мне сказать, почему?Спасибо команде optaplanner!

  Solving ended: 
    time spent (900000), 
    best score (0hard/-1764medium/0soft), 
    score calculation speed (381/sec), 
    phase total (1), 
    environment mode (REPRODUCIBLE).

Explanation of score (-1000hard/-1764medium/-200soft):
    Constraint match totals:
        -1000hard: constraint (task ETD not in holiday) has 1 matches:
            -1000hard: justifications ([Task [Task_name=1-PKGK19VE0006414A-EGV-2019-08-31/2019-08-01/SEA-3-N, extract_ct=3, waitting_time=84, searchRange=85, searchRangeDown=54]])
        -1319medium: constraint (Loading not over ots_ratio) has 10 matches:
            -390medium: justifications ([染色 12 taskList: [] ots_ratio:-390 ttl_ots_ratio:0])
            -190medium: justifications ([染色 19 taskList: [] ots_ratio:-190 ttl_ots_ratio:0])
            ...
        -445medium: constraint (Gmt_Loading not over overcapacity_ratio) has 22 matches:
            -57medium: justifications ([EAV 19-W22-(5&6) taskList: [] overcapacity_ratio:-58])
            -51medium: justifications ([EAV 19-W30-7 taskList: [] overcapacity_ratio:-52])
            ...
        -200soft: constraint (task ETD should equals old_delivery_date) has 1 matches:
            -200soft: justifications ([Task [Task_name=1-PKGK19VE0006414A-EGV-2019-08-31/2019-08-01/SEA-3-N, extract_ct=3, waitting_time=84, searchRange=85, searchRangeDown=54]])
    Indictments (top 5 of 33):
        -1000hard/-200soft: justification (Task [Task_name=1-PKGK19VE0006414A-EGV-2019-08-31/2019-08-01/SEA-3-N, extract_ct=3, waitting_time=84, searchRange=85, searchRangeDown=54]) has 2 matches:
            -1000hard: constraint (task ETD not in holiday)
            -200soft: constraint (task ETD should equals old_delivery_date)
        -390medium: justification (染色 12 taskList: [] ots_ratio:-390 ttl_ots_ratio:0) has 1 matches:
            -390medium: constraint (Loading not over ots_ratio)
        -190medium: justification (染色 19 taskList: [] ots_ratio:-190 ttl_ots_ratio:0) has 1 matches:
            -190medium: constraint (Loading not over ots_ratio)
        -164medium: justification (抓毛 12 taskList: [] ots_ratio:-164 ttl_ots_ratio:0) has 1 matches:
            -164medium: constraint (Loading not over ots_ratio)
        -127medium: justification (染色 5 taskList: [] ots_ratio:-127 ttl_ots_ratio:0) has 1 matches:
            -127medium: constraint (Loading not over ots_ratio)
        ...

ниже приведен один из слюней Жесткое ограничение:

rule "task ETD not in holiday"
   salience 300
    when 
        $Task:Task(!plan_flag.equals("1.MPS计算-N"),$Task_name:Task_name,$ETD:JS_ETD_Calendar_ETD_drools(waitting_time,extract_ct,destination,Delivery_Method),$check_holiday:check_holiday(JS_ETD_Calendar_ETD_drools(waitting_time,extract_ct,destination,Delivery_Method)));
        eval($check_holiday>0)//eval($check_holiday>0)

    then

    scoreHolder.addHardConstraintMatch(kcontext, -1000); 
    end

below in main class,I use solver.explainBestScore():
    System.out.println(solver.explainBestScore());

Answer 1

Включите <environmentMode>FULL_ASSERT</>. Вы, вероятно, страдаете от искаженной оценки, и тогда это произойдет именно тогда, когда это произойдет.