Yes you are absolutely right that the prediction of any machine learning model is based on the training data.
So if your data is constantly changing and you want to update your machine learning model as well, you can do the following steps :-
Whenever any new user's activity is added in your database table, take that data and add to your training data. For this purpose you need to plan how well you can structure the table and save your data.
Then set a cron job (cron job is time-based job scheduler that runs your script(job) on the basis of the time you set).
Set a threshold value like whenever this much activity is increased in the database or this much data is increased , run the cron job or run the script.
In your cron job, the script will do three things.
- First it will take the new data from the database , prepare it in the format of training data.
- Then train the saved machine learning model on the new data prepared in the above steps.
- After the training is finished, replace the old machine learning model with the new one.
In this way, you can easily keep your model up-to-date.
与恶龙缠斗过久,自身亦成为恶龙；凝视深渊过久,深渊将回以凝视…
