Supervised learning, regression

Question

Hi guys, I wrote this code for this question and my score is 0.10724 and the first person's score is 0.10676. I want those who know how to look at machine learning to improve and optimize my code and tell me why my score gets worse no matter what I do and even when I use a neural network model? Can someone fix my code please?......Description
🧪 Problem Description
Steel slab quality plays a crucial role in industrial manufacturing. Predicting the quality of a slab during production can lead to better decision-making, reduced waste, and more efficient processes.
📊 What You’ll Do
Explore and preprocess the dataset
Engineer or select relevant features
Train a regression model using the train.csv data
Predict the quality score of slabs in test.csv
Submit your predictions and compete on the leaderboard
📁 Dataset
train.csv: Includes features and target variable (quality_score)
test.csv: Contains features only
📈 Evaluation
Submissions are evaluated using Root Mean Squared Error (RMSE)
RMSE

Answer 1

https://sololearn.com/compiler-playground/chZ925c7OyJN/?ref=app

Answer 2

Shadow unfortunately this won't work on sololearn...

Try this:

https://sololearn.com/compiler-playground/cJBdhVo1pzRb/?ref=app

Answer 3

You can enhanced feature engineering with more datetime features, interaction terms, polynomial features, and binning for better pattern capture. Also try using multiple meta-models e.g. Lasso, Ridge, ElasticNet with blended predictions for more robust final results.

It is very hard to debugged this code in this way and sorry for you from Sololearn the code won't run here, instead your csv files has no directory.

About neural network, if your dataset is small which is common in industrial applications, neural networks may overfit. Neural networks require more tuning and may not provide enough improvement to justify the complexity.

Answer 4

Shadow ,

a very interesting task. just for my understanding:

what exactly does your data look like in the input dataset? could you possibly link the data you used here? or at least provide a description so that we can create our own dataset.

> thanks !

Answer 5

BroFar hi thanks a million

Answer 6

It seems your code and your description are from one of the contests on Kaggle. That code won't work on SL.

In order to import those different models in the code on SL, I think you need to import os and use os.system to install them, but you are also linking to a train set and a test set that don't exist on SL in that context.

It doesn't make sense trying to create that submission.csv file either on SL.

Your code probably works great on a pc with the right tools installed, such as Jupyter notebook etc, but it's quite hard trying to debug the score from the ML models when the code doesn't run on SL.

Answer 7

Thanks. I worked on this code for about 15 hours and got stuck. Can you improve my code? I did not train the model on the coolab site. `ᴴᵗᵗየ

Answer 8

Please help me