Mahout的taste推薦系統引擎（影片推薦案例）

开源遗迹發表於2024-04-04
AST
這是一個b站使用推薦引擎推薦的案例：
0點贊1收藏2轉發
<select id="getAllUserPreference" resultType="com.imooc.bilibili.domain.UserPreference">
    select
        userId,
        videoId,
        sum(case operationType
                when '0' then 6
                when '1' then 2
                when '2' then 2
                else 0 end
            ) as `value`
    from
        t_video_operation
    group by userId, videoId
</select>
public class UserPreference {

    private Long id;

    private Long userId;

    private Long videoId;

    private Float value;

    private Date createTime;

    public Long getId() {
        return id;
    }

    public void setId(Long id) {
        this.id = id;
    }

    public Long getUserId() {
        return userId;
    }

    public void setUserId(Long userId) {
        this.userId = userId;
    }

    public Long getVideoId() {
        return videoId;
    }

    public void setVideoId(Long videoId) {
        this.videoId = videoId;
    }

    public Float getValue() {
        return value;
    }

    public void setValue(Float value) {
        this.value = value;
    }

    public Date getCreateTime() {
        return createTime;
    }

    public void setCreateTime(Date createTime) {
        this.createTime = createTime;
    }
}

/**
 * 基於使用者的協同推薦
 * @param userId 使用者id
 */
public List<Video> recommend(Long userId) throws TasteException {
    List<UserPreference> list = videoDao.getAllUserPreference();
    //建立資料模型
    DataModel dataModel = this.createDataModel(list);
    //獲取使用者相似程度
    UserSimilarity similarity = new UncenteredCosineSimilarity(dataModel);
    System.out.println(similarity.userSimilarity(11, 12));
    //獲取使用者鄰居
    UserNeighborhood userNeighborhood = new NearestNUserNeighborhood(2, similarity, dataModel);
    long[] ar = userNeighborhood.getUserNeighborhood(userId);
    //構建推薦器
    Recommender recommender = new GenericUserBasedRecommender(dataModel, userNeighborhood, similarity);
    //推薦影片
    List<RecommendedItem> recommendedItems = recommender.recommend(userId, 5);
    List<Long> itemIds = recommendedItems.stream().map(RecommendedItem::getItemID).collect(Collectors.toList());
    return videoDao.batchGetVideosByIds(itemIds);
}


/**
 * 基於內容的協同推薦
 * @param userId 使用者id
 * @param itemId 參考內容id（根據該內容進行相似內容推薦）
 * @param howMany 需要推薦的數量
 */
public List<Video> recommendByItem(Long userId, Long itemId, int howMany) throws TasteException {
    List<UserPreference> list = videoDao.getAllUserPreference();
    //建立資料模型
    DataModel dataModel = this.createDataModel(list);
    //獲取內容相似程度
    ItemSimilarity similarity = new UncenteredCosineSimilarity(dataModel);
    GenericItemBasedRecommender genericItemBasedRecommender = new GenericItemBasedRecommender(dataModel, similarity);
    // 物品推薦相擬度，計算兩個物品同時出現的次數，次數越多工的相擬度越高
    List<Long> itemIds = genericItemBasedRecommender.recommendedBecause(userId, itemId, howMany)
            .stream()
            .map(RecommendedItem::getItemID)
            .collect(Collectors.toList());
    //推薦影片
    return videoDao.batchGetVideosByIds(itemIds);
}

private DataModel createDataModel(List<UserPreference> userPreferenceList) {
    FastByIDMap<PreferenceArray> fastByIdMap = new FastByIDMap<>();
    Map<Long, List<UserPreference>> map = userPreferenceList.stream().collect(Collectors.groupingBy(UserPreference::getUserId));
    Collection<List<UserPreference>> list = map.values();
    for(List<UserPreference> userPreferences : list){
        GenericPreference[] array = new GenericPreference[userPreferences.size()];
        for(int i = 0; i < userPreferences.size(); i++){
            UserPreference userPreference = userPreferences.get(i);
            GenericPreference item = new GenericPreference(userPreference.getUserId(), userPreference.getVideoId(), userPreference.getValue());
            array[i] = item;
        }
        fastByIdMap.put(array[0].getUserID(), new GenericUserPreferenceArray(Arrays.asList(array)));
    }
    return new GenericDataModel(fastByIdMap);
}

相關文章