Shulou(Shulou.com)06/03 Report--

Brief introduction:

LibRec is a leading recommendation system Java open source algorithm tool library, covering more than 70 various types of recommendation algorithms, effectively solving the two key recommendation problems of score prediction and item recommendation. Recommendation system is a classic practical application of machine learning and big data technology, which aims to provide efficient and accurate personalized item recommendation, and is an important component of modern Web applications.

The structure of the project is clear, the code style is good, the testing is sufficient, and the notes and manuals are perfect. At present, the project uses the GPL3.0 protocol to open source in github, welcome to try it out.

Librec: http://www.librec.net/

GithubRepo: https://github.com/guoguibing/librec

DocLink: http://wiki.librec.net/doku.php

Features:

L rich set of algorithms

Up to now, LibRec has integrated more than 70 recommendation algorithms of various types. Specifically, it is divided into benchmark algorithm, collaborative filtering algorithm, content-based algorithm, context awareness algorithm, hybrid algorithm and other extended algorithms. In version 2.0, more than 40 new algorithms have been added, including probability graph model, tensor decomposition model, factor decomposer, comment-based model, deep learning module (RBM) and other novel algorithms. Each core developer of the team is often responsible for the development and testing of a certain type of algorithm.

L good modularization

Compared to LibRec 1.x, the new version makes very deep optimizations in the underlying structure, especially in terms of modularity. The new version of the recommendation library can be divided into the following three parts: data preprocessing, recommendation algorithm and training post-processing. In the data preprocessing module, it mainly focuses on data conversion and segmentation. Support the input and conversion of data in two formats, one is the common User-Item-Rating format, and the other is the more general ARFF format. Users can also extend new types of data to enhance the existing ARFF format. In terms of data segmentation, it is supported by Ratio,Given-N,k-fold Cross validation, Leave-one-out and other methods. In the recommendation model module, including context awareness and algorithm integration. Context awareness refers to the situational information that the algorithm depends on, such as user similarity, and algorithm integration is the logical implementation of the algorithm. After model training, LibRec supports two operations: one is to evaluate the test set to get test results such as MAE, RMSE, AUC, MAP, NDCG, etc., and the other is to perform query operations such as rating prediction or item recommendation for a given user (or scenario). Users can customize more filtering operations by implementing the filter interface.

L flexible framework configuration

The new version of LibRec inherits configuration-based features, but has been updated and developed. The new configuration implementation refers to the implementation characteristics of other well-known data mining tool libraries, and has been effectively improved in flexibility. Specifically, we extract many common configuration items and retain specific configuration parameters for independent algorithms. In order to improve the configurability of the algorithm, we retain the available configuration settings for most algorithms.

L efficient execution performance

LibRec always pays great attention to the efficiency of algorithm execution and optimizes the framework structure and algorithm implementation as much as possible. Compared with other recommendation algorithm libraries, LibRec can be executed in a shorter time on the premise of getting considerable recommended performance.

Simple framework usage

Earlier versions of LibRec can only run independently, so it is difficult to integrate and use in other projects. Because of the good module structure, the new version can be run independently or can be used in other projects as a dependent library.

L good scalability

Good scalability. LibRec provides a good public interface for users to personalize extensions. It includes the extended interface of data type, recommendation algorithm, output type, evaluation factor, filter and so on. Using LibRec to develop new algorithms, users usually only need to focus on the logical implementation of the new algorithm, without worrying about the implementation of other parts.

The structure indicates:

Process hint:

Related links:

General knowledge of Librec2.0: https://mp.weixin.qq.com/s/AB39ihVWXYHRbeODbGO-2g

Import LibRec to Eclipse platform: https://mp.weixin.qq.com/s/OyYn5_4GYAbF0L0SFgsHVQ

LibRec command line operation: https://mp.weixin.qq.com/s/xnkg6BGyUUKmbs009p8XCw

Librec one week age: https://mp.weixin.qq.com/s/vDnca1FMW9vVrFDgti_1IA

Welcome to the official account of Librec Wechat:

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.