The Research And Design Of Personalised Job Recommender System

As for current online job platform, the overload of information puts a great negative effect on the efficiency of job hunters. To solve the problem, recom-mender systems were introduced. Recommender systems are able to reduce the effect of information overload and improve efficiency. Some scholars have done a lot of researches in the area of job recommendation and get up with some innovative algorithms and models. These algorithms and models mainly provide personalized recommendation service by getting matches between jobs and job seekers through calculating similarities and learning individual prefer-ences through interaction history of users. Even though they have accomplishedsome good results, the accuracy and the success rate of recommendation is low.Main reasons lie in the former methods can not mine data comprehensively and ignore the combination between matches of job hunters and individual pref-erence. Besides, former researches, in the area of job recommendation, focus on the improvement of recommendation performance and ignore the protection of the recommender system security. To resolve problems above, this passage design a personalized job recommender system by literature review and exper-iment verification.The work of this passage mainly focuses on the following areas:1. This passage introduced the function and value of job recommender sys-tems and gave a review on the development and current situation of job recommendations.2. To gain a better recommendation accuracy, this passage proposed a model named T-GBRT, introducing the important context information ——time factors into the job recommendation model. The effect on individual pref-erence was formalized into aggregation effect and attenuation effect and treated as a feature in the feature selection process. In the section of pref-erence learning, this passage used GBRT algorithm to train models. To improve the ability of adaption to large scale data, the passage proposeda filtering method based on neighbors. This filtering method reduced the amount of calculation by filtering mass of jobs tone recommended.3. In the section of protection of the recommender system, .this passage pro-posed a security detection model based on naive bayes classification. At-tackers would inject into the platform with many fake user ratings to inter-fere with the recommendation results. This passage treated the detection of fake users as binary classification problem. In describing user ratings,this passage introduced normalized biased ratings to describe the bias be-tween a single user rating and overall ratings more comprehensively.4. Designing experiment and verify the performance of proposed models.5. Give the overall design of the personalized job recommender system, in-cluding user interaction model, data collection and processing section, rec-ommendation engine section, data storage section and security detection section. On the basis of improving and perfecting designs, the passage gave the realization of the design.By experiment verification, the personalized job recommender system proposed in this passage achieved expected goals on the improvement of recommendation accuracy and security detection.