{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/c7d9b89a54234f699204b16a9a313c7d\" frameborder=\"0\" width=\"1280\" height=\"960\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":960,"width":1280,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":960,"thumbnail_width":1280,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/c7d9b89a54234f699204b16a9a313c7d-df9ba0c970f88f11.gif","duration":300.172,"title":"Predicting LinkedIn Job Post Views","description":"This Loom presents Mifla Kesselman’s machine learning analysis of LinkedIn job postings to predict engagement. She sampled 30,000 postings from about 123,000 total records, transforming features such as salary and transparency into binary variables and applying a log transform to views due to a right-skewed distribution. In regression, the tuned Random Forest model was the best but explained only about 8% of the variance, suggesting most variation is driven by LinkedIn’s unseen algorithm. For classification of high engagement versus normal, she trained four models and found a decision tree performed best with the highest F1 and recall for the high-engagement class. She also used clustering to create additional features and showed exploratory findings like higher average views for certain days and weak linear correlation overall."}