Our client, Socialinsider, provides social media analytics tools in measuring and comparing social media performances for business. We help our client by improving their marketing strategy to better convert subscribed users and expanding their product by predicting future social media performance metrics.
To improve the conversion rate of subscribed users, we worked with web log data that each row represents an online log, with columns include user id, the length they stay on the page, and any type of action users have. We first did exploratory analysis and then did feature engineering according to the insights. We dealt with data imbalance with upsampling and downsampling as the conversion rate is very low. We tried, compared, and fined tuned different models and select gradient boost model as our final model. In the end, we deliver the model with recall of 87%, which exceeds the industry benchmark. Our client apply our model on their weekly meetings.
Our team developed a time-series forecasting pipeline that processes and cleans historical follower count data from social media profiles. We use linear regression as a baseline model, and decided to use AutoTS, an automated time-series modeling tool, as the final model since it gets the best result. The model was trained using historical follower trends and evaluated on unseen data. Key evaluation metrics such as Mean Absolute Error (MAE) and Mean Squared Error (MSE) were used to assess prediction accuracy. The trained model generates a future prediction table with projected follower counts, which can be used by marketing teams to set realistic expectations and optimize content strategies. For socialinsider’s benefits, the deliverable will allow forecasts for individual profiles.
Our team also predicted the average engagement per post for the next week for each Instagram account. We incorporate post metrics such as reach, impression, and followers, and lag features of average engagement from previous week and month. The model reach adjusted R^2 of 0.86.
Mentor: Alessio Brini
