This article is reproduced in https://blog.csdn.net/u011095110/article/details/84403564
1. Overview of recommended algorithms
1) based on recommendations: some knowledge of this type are generally dependent on natural language processing NLP by TF-IDF text mining feature vectors to get the user's preferences, and then make recommendations. Such a recommendation algorithm to find a unique niche preferences of the user, but also a better explanatory. Because this type of NLP needed basis, this article is not to talk, specifically about NLP discuss in a later time.
2) Coordination filtering recommendation: later in this article to be devoted to content. Coordination filtering recommendation algorithm is the most mainstream type, variety, has been widely used in many industries. The advantage is that does not require much knowledge in specific areas, you can get a better recommendation results by machine learning algorithms based on statistics. The biggest advantage is easily achieved on the project, it can be easily applied to products. Currently the vast majority of the practical application of the recommendation algorithm is collaborative filtering algorithms.
3) hybrid recommendation: This is similar to our machine learning ensemble learning, Bo long before the public, by combining multiple recommendation algorithm to get a better recommendation algorithms, Three Stooges play a top role of Zhuge Liang. For example, by creating multiple models recommendation algorithm, and finally voting method to determine the final recommendation result. Hybrid recommendation theoretically not worse than any single recommendation algorithm, but using hybrid recommendation algorithm complexity would be increased, there is use in practical applications, but there is no single coordinating filtering algorithm, such as logistic regression and the like wide binary classification recommendation algorithm.
4) rule-based recommendation: This type of algorithm based on the most common such as user clicks, the user browsing the most recommended method belonging to the mass, and in the current era of big data is not mainstream.
5) recommendation based on demographic information: This category is the simplest recommendation algorithms, and it simply found that the degree of correlation based on the user's basic information system users, and then make a recommendation, now rarely used in large systems
2. Recommended overview of collaborative filtering
Collaborative filtering (Collaborative Filtering) as recommended in the most classic type of algorithm, including online and offline collaborative filtering in two parts. The so-called online collaboration, is to find a user might like items through online data, offline filtration, while some are not recommended data filter out, than such recommended low score of the data, or it is recommended high but the user has purchased data .
Collaborative filtering model is generally items of m, m user data between the user and part of the data is only part of the score data, other parts of the score blank, at this time we have to use part of the sparse data to predict those ratings relationship between objects and data blank, find the highest score of recommended items to the user.
In general, collaborative filtering is divided into three types. The first is based on the user (user-based) collaborative filtering, a second item (item-based) based on collaborative filtering, and the third is based on a model (model based) collaborative filtering.
Based on the user (user-based) collaborative filtering the main consideration is the similarity between the user and the user, as long as users like to find similar items, and to predict the target user score corresponding to the item, you can find a number of top-rated items recommended to the user. The project-based (item-based) collaborative filtering and user-based collaborative filtering is similar to, but then we turn to find the similarity between the goods and items, only to find the target user rating of certain items, then we We can predict the high similarity of similar items, the highest score of a number of similar items recommended to the user. For example, you bought online a machine learning-related books, the site will soon recommend a bunch of machine learning, large data relevant to your book, here obviously uses collaborative filtering based on the idea of the project.
High user-based collaborative filtering need to find the similarity between the user and the user is online, the computational complexity would surely be based on project-based collaborative filtering: simply compare our user-based collaborative filtering and item-based collaborative filtering. But it can help users find new category of surprise items. The collaborative filtering-based project, due to the similarity of items to consider for some time does not change, so you can easily off-line calculation accuracy generally acceptable, but it is recommended diversity, it is difficult to give the user a pleasant surprise a. For small general recommendation systems, collaborative filtering-based projects is definitely mainstream. However, if the recommendation is a large system, you can consider user-based collaborative filtering, of course, more can consider our third type, model-based collaborative filtering.
Based on the model (model based) collaborative filtering is the most mainstream type of collaborative filtering, and a lot of our machine-learning algorithm may also find their way here. Here we focus on model-based collaborative filtering.
3. The model-based collaborative filtering
With the idea of machine learning to solve the model, the mainstream method can be divided into: a correlation algorithm, clustering, classification algorithms, regression algorithms, matrix decomposition, neural networks, graphical models and semantic model to address hidden
3.1 associated with doing collaborative filtering algorithm
Generally, we can identify the item sets live sequence data for all items purchased by the user in frequent, do frequent set mining, N find frequent item sets or sequences to meet the support threshold associated items. If you purchase the frequent N item sets or sequences where some of the items, then we can be frequent item sets or sequences in other items recommended by a certain rating criteria to the user, the scoring criteria may include support, confidence and lift and so on . Commonly associated recommendation algorithm Apriori, FP Tree and PrefixSpan
3.2 do collaborative filtering algorithm with clustering
3.3 classification algorithm used to do collaborative filtering
Logistic regression and naive Bayes algorithm
3.4 regression algorithm to do collaborative filtering
Linear regression and decision tree algorithm and Support Vector Machine
3.5 do collaborative filtering matrix factorization
Disassembler (Factorization Machine) and tensor decomposition (Tensor Factorization)
3.6 do collaborative filtering using neural network
CNN, RBM and RNN
3.7 do collaborative filtering model with FIG.
SimRank系列算法和马尔科夫模型算法
3.8用隐语义模型做协同过滤
基于NLP的，主要方法有隐性语义分析LSA和隐含狄利克雷分布LDA