Predicting Fake Job Posts
Download as PPTX, PDF1 like1,385 views
This document compares the performance of two machine learning algorithms, two-class boosted decision tree and two-class decision forest, in predicting fake job postings. The boosted decision tree algorithm achieved higher accuracy (93.8% vs 95.4%) and recall (75% vs 2%) than the decision forest algorithm, suggesting it is better able to detect fake job postings. While the decision forest algorithm had perfect precision of 100%, the boosted decision tree provides a more reliable overall performance in identifying fake job postings.
Predicting Fake Job Posts
- 1. Performance comparison of Two class boosted decision tree and Two class decision forest algorithms in predicting fake job postings FHA. Shibly1, Uzzal Sharma2 and HMM. Naleer3 Presented By :- FHA. Shibly Senior Lecturer, South Eastern University of Sri Lanka Ph.D. Candidate, Assam Don Bosco University, India
- 2. Introduction – Fake Job Postings Many businesses prefer to post their vacancies electronically so that job applicants can access them quickly and timely. But this purpose may be one form of scam on the part of the fraud individuals because they give job applicants during terms of taking money from them or collecting their personal information for involving in cybercrimes. Fake job posting advertisements can be written against a reputable firm for breaching its reputation. The fraudulent post-detection work draws proper attention to obtaining an automated tool to identify fake jobs and report them to people to avoid applying for such situations. At present, many machine learning algorithms have been used to detect such fraudulent posts. But, the performance of such algorithms to be measured and compared to find a proper algorithm to incorporate in identifying fake things
- 3. Introduction – Machine Learning • Two class boosted decision tree and Two class decision forest algorithms are widely used. A boosted decision tree is an ensemble learning method in which the second tree corrects for the errors of the first tree, the third tree corrects for the errors of the first and second trees, and so forth. Predictions are based on the entire ensemble of trees together that makes the prediction. • Decision forests are fast, supervised ensemble models. This module is a good choice if you want to predict a target with a maximum of two outcomes.
- 4. Methodology • Dataset • Shivam has provided a data set containing 18 000 job descriptions, of whom approximately 800 are fake. • MS Azure machine learning studio Dataset Algorithms Split Data Model Training Model Scoring Model Evaluation
- 5. Results Two class boosted decision tree Two class decision forest algorithms Accuracy 0.938 0.954 Precision 0.720 1.000 Recall 0.750 0.020 F1 Score 0.735 0.039
- 6. Results - Receiver Operating Characteristic curve
- 7. Conclusion • Researchers tried to measure the efficiency of two-class boosted decision tree and two-class decision forest algorithms in predicting fake job postings by using the proposed model. • The findings and results showed that two-class boosted decision tree is healthier than the Two class decision forest algorithms in detecting fake job posts. • Therefore, algorithm 01 can be used in finding and identifying fake or rumor posts, comments and publications in social media. It will return more reliable outputs than algorithm 02. In the future, more algorithms can be tested and compared to find more reliable parameters to detect fake things to control unnecessary burdens to social media users. The dataset and context also can be enlarged and enhanced to find more results in different approaches. Therefore, we can use and build new models by using two-class boosted decision tree to find or detect fraudulent posts specially job postings in social media or digital pages.