I am trying to model a tourist satisfaction model based on multicriteria. I want to use a supervised machine learning algorithm. In this regard, I want to label my data using a satisfaction equation based on the HOLSAT model and later test and train the labeled data set using a machine learning algorithm. Is this data labeling procedure good enough for my work? Please suggest?
I have no idea about HOLSAT model. But you can do data labeling by using any supervise classification algorithm (e.g. SVM, RF, XGBoost). In this case, you have to convert your training data label into binary class (e.g. one hot encoding in Python or factorization in R). Then, you can use the training model to classify test dataset. Finally, you can inverse your binary encoding values into label class.
If you also make labels based on ML, you're increasing the uncertainty in your network. Therefore, the situation is very complex and you should use several uncertainty propagations through your network.
Using an expression to label data is not anything good for the model, because if that worked, there was no need for Machine Learning and prediction. You can look into this article to start digging:
https://www.phdata.io/blog/techniques-for-labeling-data-in-machine-learning/
Dear Muntasir Hoq, Thank you for your answer. I just wanted to ask you one more thing, I am using three features only for labeling the data, whereas the whole dataset has 77 features (both categorical and numeric) and for machine learning, I want to pass all 77 features for prediction purposes. Do you still think using an equation is not a good idea for labeling? Also, how can I then proceed? Thank you once again.
- Badges
- Science topic