Job recommendation algorithm for graduates based on personalized preference

Abstract

It is challenging for graduates to find a proper job. Unlike those with occupational history, graduates generally are short of work experience and the support from social network, so they have to face hundreds of recruitment companies. The process of applying for a job is time-consuming, especially in preparing and attending tests and interviews. Not knowing which companies are most proper for them, graduates need to devote their energy and time to preparing for each potential recruitment. This job-hunting strategy can easily lead to employment dissatisfaction or failure. Therefore, it is very helpful to recommend a few most suitable jobs to graduates. Collaborative filtering (CF) method is currently the most frequently adopted and effective recommendation algorithm, but it cannot be directly applied to job recommendation for graduates because graduates generally have no historical records on employment. Besides, job recommendation should take into account graduate preferences for jobs, such as enterprise types and company locations, which are crucial to job choices. To address these challenges, we first analyze the pattern of job choices of graduates. Based on this, we propose a personalized preference collaborative filtering recommendation algorithm (P2CF), which can not only recommend jobs for graduates through massive campus records, but also identify graduate personal preferences for jobs. Graduates are first clustered into different groups according to their academic performances and family economic conditions. Then Bayesian personalized ranking (BPR) method is introduced to calculate the scores of graduate groups to jobs. Finally the scores and graduate personalized preferences are combined to recommend a few potential jobs. P2CF is a recommendation algorithm with hierarchical structure, which takes account of both the group records of job choices and the individual preferences for jobs. Experimental results show that P2CF on job recommendation outperforms state-of-the-art CF methods and identifies graduate personalized preference for jobs accurately.

A data mining approach to forecast students’ career placement probabilities and recommendations in the programming field

“…The performance of the classifiers is then compared in terms of accuracy [35]. Among the classifiers, the Classification Tree gave the highest accuracy for both datasets.…”

“…The data was divided among the 14 faculties with 61271 undergraduate students per 2389 courses. This record contains 35 fields with details about students like grades in some courses. Still, the study focused on four primary areas like student identity, faculty name, course identification, and related courses.…”

Web Based Online Hybrid Teaching Method of Network Music Course

“…Das and other scholars proposed a collaborative filtering method based on clustering, which uses two hierarchical spaces to divide the data. The experimental results show that this method not only improves the operation speed, but also ensures the recommendation quality [7]. Wang's team used the improved factor decomposition machine model to design a new hierarchical decomposition machine, and used the generalized Kalman filter and expectation maximization algorithm for recommendation.…”

“…The whole algorithm is represented by spatial vector model based on knowledge base concept map and learning object. Assuming that the concept set is C, the constraint relationship set between concepts is R, and there are N different concepts in set C, set C is expressed as formula (7). .…”

Research on Product Design Strategy Based on User Preference and Machine Learning Intelligent Recommendation

“…The recommendation algorithm based on collaborative filtering does not need detailed content. When the details of content cannot be accessed or it is difficult to collect or analyze the details, collaborative filtering method is very effective, and this method can find the items that target users want in a large number of items [15]. However, it will also face the problem of rating sparsity, and there will also be the problem of cold start of new users and new projects.…”