Are there any ways to create a deep multilayer perceptron model that is capable of making accurate regression predictions based on the training done using around 1000 unique data?

I'm currently working on a Kaggle challenge for predicting the amount Followers gained using the top 1000 streamers on Twitch 2020 dataset.
- The X value would be every columns excluding Followers gained;
- The y value would be the amount of Followers gained - the model will make predictions regarding this.

In general, the data values for the amount of followers gained contain around 6 to 7 digits; currently my RMSE loss value is almost near to become 5 digits, yet still 6.
There are limited quantity of data; I'm aiming for a 5 digit RMSE value.

Here's an overview structure of my MLP model; this one showed the best result by so far. Do let me know if you have any recommendations. Thanks. enter image description here


Looks the model is good,
If training accuracy is 100% then try increasing the dropout percentage in the initial layers or reduce some hidden layers. If training accuracy is still less than 100% then try decreasing the dropout percentage in the last few layers of nodes with count as 32.
Refer this page https://stats.stackexchange.com/questions/417055/dropping-outliers-based-on-2-5-times-the-rmse you may get more insights. Thanks


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.