SOCIAL MEDIA PREDICTION

We try to predict the number of likes associated with a certain social media profile. This is a regression problem. Everything was run on Kaggle.

STRUCTURE

The datasets (raw and pre-processed) can be found here. The data exploratory analysis as well as the pre-processing can be found in the following notebook. The pre-trained models can be found here. The saved hyper-parameters for each model with the CV result can be found here. The predictions as csv can be found here.

HOW TO RUN

USING PRE-TRAINED MODELS

1. Open the predict script.
2. Based on the models dictionary in that script, change the NAME variable to reflect the model you want to use.
3. Execute the script.
4. The results are saved here

TRAIN THE MODEL FROM SCRATCH

1. Inside the training folder, execute the model you want to train by choosing the file "modelname_training.py".
2. The trained model will be saved here
3. Follow steps 1 through 4 of section "USING PRE_TRAINED MODELS".