2026年4月10日 AI中考填报助手：推荐算法与系统架构深度剖析

 传统填报方式——纯人工查询比对
def traditional_application(scores, preferences):
    schools = get_all_schools()       逐个翻看学校招生简章
    suitable = []
    for school in schools:
        if scores >= school.min_score:       只看分数是否达标
            if school.city == preferences.city:
                suitable.append(school)
     弊端：只看分数和城市，忽略历年录取波动、梯度风险、兴趣匹配
    return suitable

 简化版协同过滤：基于用户的相似度计算
import numpy as np
from sklearn.metrics.pairwise import cosine_similarity

class CollaborativeFilter:
    def __init__(self):
         user_school_matrix: 行=用户, 列=学校, 值=录取结果(1录取/0未录)
        self.user_school_matrix = None
    
    def fit(self, matrix):
        """输入用户-学校交互矩阵"""
        self.user_school_matrix = matrix
    
    def find_similar_users(self, target_user_id, top_k=5):
        """找到与目标用户最相似的K个历史考生"""
         使用余弦相似度计算用户之间的相似程度
        similarities = cosine_similarity(
            self.user_school_matrix[target_user_id:target_user_id+1],
            self.user_school_matrix
        )[0]
         排除自身，按相似度排序
        similar_indices = np.argsort(similarities)[::-1][1:top_k+1]
        return similar_indices, similarities[similar_indices]
    
    def recommend_schools(self, target_user_id, top_n=10):
        """基于相似用户的录取记录进行推荐"""
        similar_users, scores = self.find_similar_users(target_user_id)
         聚合相似用户录取的学校，按出现频率加权
        recommendations = {}
        for user, sim_score in zip(similar_users, scores):
            user_schools = self.user_school_matrix[user]
            for school_id, admitted in enumerate(user_schools):
                if admitted == 1:   该用户录取了这所学校
                    recommendations[school_id] = recommendations.get(school_id, 0) + sim_score
         返回推荐度最高的Top N所学校
        return sorted(recommendations.items(), key=lambda x: x[1], reverse=True)[:top_n]

 使用示例
 cf = CollaborativeFilter()
 cf.fit(user_school_matrix)    构建用户-学校矩阵
 recommendations = cf.recommend_schools(target_user_id=999)   为新考生推荐学校

 简化版内容推荐：基于学校特征向量的匹配度计算
class ContentBasedRecommender:
    def __init__(self):
         school_features: 每所学校的特征向量
         维度示例：[录取分数线, 一本率, 师资评分, 社团丰富度, 地理位置评分]
        self.school_features = None
    
    def fit(self, features):
        """输入学校特征矩阵"""
        self.school_features = features
    
    def get_user_profile(self, target_scores, preferences):
        """根据考生成绩和偏好构建用户画像向量"""
         维度对齐：分数线期望、升学质量要求、兴趣匹配等
        profile = [
            target_scores,               成绩位次
            preferences.get('升学率权重', 0.7),
            preferences.get('社团权重', 0.3),
            preferences.get('地理偏好权重', 0.5)
        ]
        return np.array(profile)
    
    def recommend(self, user_profile, top_n=10):
        """计算每所学校与用户画像的匹配度"""
         欧氏距离或余弦相似度
        similarities = []
        for i, school_vec in enumerate(self.school_features):
             简化版：点积作为匹配分数
            score = np.dot(user_profile, school_vec[:len(user_profile)])
            similarities.append((i, score))
        return sorted(similarities, key=lambda x: x[1], reverse=True)[:top_n]

传统填报：经验判断 → 信息碎片化 → 决策风险高
     ↓
AI填报：协同过滤(群体经验) + 内容推荐(个体偏好) → 混合推荐
     ↓
输出：“冲、稳、保”四梯度志愿方案 + 录取概率预估

 方式一：传统方式——纯经验判断
def traditional_recommend(scores):
     简单粗暴：分数够了就推荐，完全不考虑位次、梯度、兴趣
    if scores >= 680:
        return ["一中", "二中"]
    elif scores >= 650:
        return ["三中", "四中"]
    else:
        return ["普高A", "普高B"]
 弊端：一刀切、缺乏个性化、无法应对每年录取波动

 方式二：基于协同过滤（CF）的推荐
def cf_recommend(user_id, user_school_matrix):
    cf = CollaborativeFilter()
    cf.fit(user_school_matrix)
    return cf.recommend_schools(user_id, top_n=10)
 优势：利用历史录取数据，推荐更贴合实际录取情况
 局限：新考生冷启动困难，无法处理兴趣偏好

 方式三：混合推荐（Hybrid）——当前主流方案
def hybrid_recommend(user_id, scores, preferences, user_school_matrix, school_features):
     Step1: 协同过滤推荐（基于历史录取模式）
    cf = CollaborativeFilter()
    cf.fit(user_school_matrix)
    cf_recs = cf.recommend_schools(user_id, top_n=20)
    
     Step2: 内容推荐（基于考生个体画像）
    cb = ContentBasedRecommender()
    cb.fit(school_features)
    user_profile = cb.get_user_profile(scores, preferences)
    cb_recs = cb.recommend(user_profile, top_n=20)
    
     Step3: 融合排序（加权组合）
    hybrid_scores = {}
    for school_id, score in cf_recs:
        hybrid_scores[school_id] = hybrid_scores.get(school_id, 0) + score  0.6   CF权重60%
    for school_id, score in cb_recs:
        hybrid_scores[school_id] = hybrid_scores.get(school_id, 0) + score  0.4   CB权重40%
    
     Step4: 按梯度分类输出“冲、稳、保”方案
    return classify_by_gradient(hybrid_scores, scores)

def classify_by_gradient(scores_dict, user_score):
    """将推荐学校按录取难度划分为冲、稳、保三档"""
    result = {"冲": [], "稳": [], "保": []}
    for school_id, match_score in scores_dict.items():
        expected_min_score = get_school_min_score(school_id)   获取该校历年最低录取线
        diff = expected_min_score - user_score
        if diff > 15:
            result["冲"].append(school_id)    差距较大但有机会
        elif diff > -10:
            result["稳"].append(school_id)    匹配度高，稳妥选择
        else:
            result["保"].append(school_id)    远低于用户成绩，保底
    return result