Predicting-the-next-app-to-use.
This is a LSTM(Long-Short Term Memory)-Recurrent neural network model that helps to predict the next app a user might use based on his previous app usage data. The Dataset contains around 2000 datapoints reflecting all the applications that were used by a user within 1 week.
Since only 36 apps were used by the user in that particular week, the prediction is limited to the 36 apps.
To reuse the already trained model:
python src/generateModel.py True
To generate and train a new model:
python src/generateModel.py
The same "dataset.csv" is divided into training and test data. Since our mobile phone's usually contain around 4 app suggestion space, we take the first 4(with the highest probability) predicted apps ,to test the accuracy of our model.
"actual_app_used" variable is a dataframe that contains the actual apps that were used by the user. And "prediction" variable is a dataframe that contains the predicted apps of the model during the same instance.
Final Output-
