Hi Izdiar, the process you have described for training a machine learning model is generally correct. However, I can recommend some considerations:

1) Data normalization: It is very important to normalize the numerical features in the training and testing data so that they have mean 0 and variance 1. This improves the performance and stability of the model.

2) Data division: The data set should be randomly divided into training set (generally 80%) and test set (20%). The model only sees the training set during learning.

3) Performance measurement: Metrics such as precision, recall, ROC curve should be measured on the test set to evaluate the actual performance of the trained model.

4) Model export: Once the trained model satisfies desired metrics on the test set, it is exported and saved to make inferences on new data.

5) Preprocessing of new data: Any new data that enters the already trained model must be normalized in the same way as the training data so that the characteristics are comparable.

6) Inference on new data: The new normalized data is passed to the trained model to generate predictions and evaluations.

Thank Kevin,

please could you have any example about it.

I am new in this field.

There are many options depending on the specific goal and data you are working with. Since you mention that you are new to this field, I would recommend starting with some simple examples using public data sets. If you are going to use Python, there are many examples available using libraries like Scikit-Learn, keras, pytorch and datasets like Iris, Boston Housing, MNIST (handwritten digits) etc. These are well documented. In MATLAB, there are also many built-in examples of predictive models and classifiers trained on different types of data.

It depends on the language/environment you plan to use, the field of application, and the type of task (regression, classification, clustering, etc).