Generateweightingfile.py: merge user data with normalized work year data. Generate_train_test_set.py: generate train and test set N-grams.py: generate data based-on bigram or trigram The LinkedIn profile is combined with following content:Ĥ.six parts of work experience(combined with job title, company name, company type, work duration, company location)Ħ.Three parts of other education backgroundĭatafilter.py: filter the data into two different dataset based on the aim of the recommendation (users who are relevanat and users who are not relevant)Ĭalculate_data_job_now.py: calculate user's work year of past work experience using regular experssionĭatanormalize.py: normalize results of the work year dataīag-of-words.py: generate data based on bag-of-words In order to make the robustness and scalability of the system, we use the Objected-Oriented (OO based) programming in our system, and the organization is shown in the linkedindata_old.py and the organization of the user profile is: We decided to use six past work experience, four parts of education (except university name, end date of education), skills, language. Thus, based on the statistic data of the job position, education background and skills. Summary of the user data and selecting of user profile attributes: more than %56 of users have more than 2 education background.Īverage number of the skills per user is: 19.72 And more than 48% of users have 2 job experience.Īverage number of effective education per user:īased on the statistic data, nearly 89% of the user have at least 1 education background. Thus, based on the data, we will mainly select job positions from those areas.Īverage number of the effective work experience per user:īased on the statistic data, nearly 82% of the user have at least 1 past job experience.
Major, end date of education, education details)Ĭonnections number of user: the number shows the connections of other LinkedIn users for every user.īased on the statistic data, there are mainly three kinds of job position: technical position (engineer), management position(manager) and academic position(professor). Four parts of educational backgrounds (university name, degree of education,.With job title, company name, company type, work duration, company location) Seven parts of work experience (current and past work experience, combined.The LinkedIn user profile data is a 261.2 MB CSV file with 158096 LinkedIn user profile.įor every user profile, they have 67 attributes that can be categorized as following: LinkedIn Profile Data Overview and Data Analysis This project is for Hybrid recommedation for talents, we used different combinations of different technologies such as feature representation approachs(N-Grams, Word Embedding) and different learning algorithms(SVM, Logistic regression) to find the best approach for expert recommendaion of LinkedIn user profiles. Hybrid-Talent-Recommendation-of-LinkedIn Aim
0 kommentar(er)
