Issue
The client, a leading content marketer in Finance industry wanted insights for their clients (advertisers) to target users based on interest (visits) and engagement (clicks). To provide these insights we built a solution that would predict trending topics in the future and help client’s content writers publish relevant content
Our Approach
- Scraping content from the web for over a million URLs from the client’s publishing platform
- Classifying content into 45 topics (Stocks, Bonds and Cryptocurrency) using LDA (Latent Dirichlet Allocation) algorithm
- Assigning topic to the article based on the weightage of keywords in the content
- Predicting user interest by forecasting users, visits and clicks using advanced predictive models (S-ARIMA, LSTM RNN) by topic to help
Impact
- Topic modelling strategy showed better accuracy than the client’s existing topic classification model and also a 3rd party vendor Diffbot
- Our predictive model results helped client’s content writers to publish relevant content on trending topics and take effective data informed decisions
