R的recommenderlab包可以实现协同过滤算法。这个包中有许多关于推荐算法建立、处理及可视化的函数。
本文任务：选用recommenderlab包中内置的MovieLense数据集进行分析，该数据集收集了网站MovieLens（movielens.umn.edu）从1997年9月19日到1998年4月22日的数据，包括943名用户对1664部电影的评分。
library(recommenderlab)library(ggplot2)
02. 数据处理与探索性
data(MovieLense)image(MovieLense)
# 获取评分ratings.movie <- data.frame(ratings = getRatings(MovieLense))summary(ratings.movie$ratings) ggplot(ratings.movie, aes(x = ratings)) + geom_histogram(fill = "beige", color = "black", binwidth = 1, alpha = 0.7) + xlab("rating") + ylab("count")
利用summary()获取评分数据，可知最大值为5，最小值为1，平均值为3.53。并将其柱状图进行绘制，如下所示。
![)1XC93KXGWJNOOJ`QIO]6[Q.png](https://ucc.alicdn.com/pic/developer-ecology/24cb42b574cf4a9bbe389d1c143f6790.png?x-oss-process=image/resize,w_1400/format,webp ")1XC93KXGWJNOOJ`QIO]6[Q.png")
数据标准化：在进行数据分析前，利用normalize()我们将数据进行标准化，并进行绘制。
atings.movie1 <- data.frame(ratings = getRatings(normalize(MovieLense, method = "Z-score")))summary(ratings.movie1$ratings)## Min. 1st Qu. Median Mean 3rd Qu. Max.## -4.852 -0.647 0.108 0.000 0.751 4.128ggplot(ratings.movie1, aes(x = ratings)) + geom_histogram(fill = "beige", color = "black", alpha = 0.7) + xlab("rating") + ylab("count")
用户的电影点评数：我们还对用户的电影点评数进行描述性分析，具体结果如下所示。
movie.count <- data.frame(count = rowCounts(MovieLense))ggplot(movie.count, aes(x = count)) + geom_histogram(fill = "beige", color = "black", alpha = 0.7) + xlab("counts of users") + ylab("counts of movies rated")
rating.mean <- data.frame(rating = colMeans(MovieLense))ggplot(rating.mean, aes(x = rating)) + geom_histogram(fill = "beige", color = "black", alpha = 0.7) + xlab("rating") + ylab("counts of movies ")
![G6Z%9PAO3Y%DLE]IUE(9[JS.png](https://ucc.alicdn.com/pic/developer-ecology/54e5ff7300af4944b56f5e7f48447067.png?x-oss-process=image/resize,w_1400/format,webp "G6Z%9PAO3Y%DLE]IUE(9[JS.png")
03. 建立推荐模型与模型评估
对于realRatingMatrix有六种方法：IBCF(基于物品的推荐)、UBCF（基于用户的推荐）、SVD（矩阵因子化）、PCA（主成分分析）、 RANDOM（随机推荐）、POPULAR（基于流行度的推荐）。
模型评估主要使用：recommenderlab包中自带的评估方案，对应的函数是evaluationScheme，能够设置采用n-fold交叉验证还是简单的training/train分开验证，本文采用后一种方法，即将数据集简单分为training和test，在training训练模型，然后在test上评估。接下来我们使用三种不同技术进行构建推荐系统，并利用评估方案比较三种技术的好坏。
library(recommenderlab)data(MovieLense)scheme <- evaluationScheme(MovieLense, method = "split", train = 0.9, k = 1, given = 10, goodRating = 4)algorithms <- list(popular = list(name = "POPULAR", param = list(normalize = "Z-score")), ubcf = list(name = "UBCF", param = list(normalize = "Z-score", method = "Cosine",nn = 25, minRating = 3)), ibcf = list(name = "IBCF", param = list(normalize = "Z-score")))results <- evaluate(scheme, algorithms, n = c(1, 3, 5, 10, 15, 20))
![$49`W@_WD7C]WZF~~GC)C@7.png](https://ucc.alicdn.com/pic/developer-ecology/462cf5f65d864aa0bd3e0e32f961ac9b.png?x-oss-process=image/resize,w_1400/format,webp "$49`W@_WD7C]WZF~~GC)C@7.png")
plot(results, annotate = 1:3, legend = "topleft") #ROC
![]PBY~J16F2}MRRP}SN1BKWB.png](https://ucc.alicdn.com/pic/developer-ecology/a6a14342a34242709bef1957aa97c544.png?x-oss-process=image/resize,w_1400/format,webp "]PBY~J16F2}MRRP}SN1BKWB.png")
plot(results, "prec/rec", annotate = 3)#precision-recall
按照评价方案建立推荐模型
model.popular <- Recommender(getData(scheme, "train"), method = "POPULAR")model.ibcf <- Recommender(getData(scheme, "train"), method = "IBCF")model.ubcf <- Recommender(getData(scheme, "train"), method = "UBCF")# 对推荐模型进行预测predict.popular <- predict(model.popular, getData(scheme, "known"), type = "ratings")predict.ibcf <- predict(model.ibcf, getData(scheme, "known"), type = "ratings")predict.ubcf <- predict(model.ubcf, getData(scheme, "known"), type = "ratings")# 做误差的计算predict.err <- rbind(calcPredictionAccuracy(predict.popular, getData(scheme, "unknown")),calcPredictionAccuracy(predict.ubcf, getData(scheme, "unknown")), calcPredictionAccuracy(predict.ibcf,getData(scheme, "unknown")))rownames(predict.err) <- c("POPULAR", "UBCF", "IBCF")predict.err
| RMSE | MSE | MAE | |
| POPULAR | 1.046 | 1.095 | 0.8315 |
| UBCF | 1.217 | 1.481 | 0.9662 |
| IBCF | 1.693 | 2.866 | 1.2397 |
通过结果我们可以看到：基于流行度推荐系统对于本案例数据的效果最好，RMSE，MSE，MAE都是三者中的最小值。其次是基于用户的推荐，最后是基于项目协同过滤。
4. 参考资料
1. Recommenderlab包实现电影评分预测(R语言)
2. R语言：recommenderlab包的总结与应用案例
3. recommender system handbook
4. Item-Based Collaborative Filtering Recommendation Algorithms
5. recommenderlab: A Framework for Developing and Testing Recommendation Algorithms
