Mahmut

We were discussing MCDM not AI or ML

You say" Generally, in AI and ML applications, Min-Max or Z-score data transformation techniques are frequently preferred for data normalization"

May I remind you that data normalizaction is also data transformation?

I agree with you that normalization to be used depends on data and purpose

"However, when there are alternative solutions such as Vector, Max, Logarithmic, Sum, why are these two enough?"

Because not all problems look for the same answer

a) You may have data with known upper and lo

wer bounds

b) Your data may or not be uniformely distributed

I would suggest reading thie article https://dataaspirant.com/data-normalization-techniques

The reason Min-Max scaling and Z-Score normalization are frequently used in AI/ML applications is that they are simple, easy to implement, and generally effective in many scenarios. Moreover, many machine learning algorithms are not highly sensitive to the specific normalization technique used, and these methods often work well as a starting point. Additionally, there are other scaling techniques available, such as Robust Scaling, which is less sensitive to outliers compared to Min-Max scaling and Z-Score normalization.

Different datasets and algorithms may benefit from different normalization strategies. For example, in deep learning models, batch normalization is a widely used technique that normalizes the activations of each layer to accelerate training and improve convergence.

Dear Suraj, First of all, thanks for your reply.

In your opinion, "Min-Max scaling and Z-Score normalization are often used in AI/ML applications because they are simple, easy to implement, and often effective in many scenarios". But Max, Vector, Sum, or Logarithmic normalization techniques are also simple and easy to implement. These methods can also be effective in different scenarios.

Secondly, again according to you "Also, many machine learning algorithms are not very sensitive to the particular normalization technique used and these methods usually work well as a starting point". Yes, several types of ML algorithms may not need normalization. But usually, normalization is used in many other methods and seen as a necessity. For example, for a house price prediction, there are many features (criteria) in different units and these are normalized for many ML algorithms.

Again in your view "There are other scaling techniques such as Min-Max scaling and Robust Scaling which are less sensitive to outliers than Z-Score normalization". In fact, apart from these two, there are at least 15-20 other normalization techniques.

I agree with what you said yes "different datasets and algorithms can benefit from different normalization strategies". But it is generally not used and Min-Max and Z-Score are generally used. This is a choice open to criticism. In my view, you have to take into account that some types of normalization may yield different results, especially in cases of decision-making among alternatives for ML.

Dear Mahmut,

Thank you for your thoughtful response. I appreciate your insights on different normalization techniques in AI/ML applications.

You are absolutely right that there are several other normalization techniques, such as Max, Vector, Sum, and Logarithmic normalization, which are simple to implement and can be effective in various scenarios. These techniques should indeed be considered when dealing with specific datasets and application contexts. Each normalization method has its advantages and may better suit certain data distributions and algorithm requirements.

Regarding your point about normalization being necessary in many ML algorithms, I agree that it is often essential to bring features to a similar scale for effective model training and to avoid potential bias towards features with larger scales. As you mentioned, in the context of house price prediction with various features measured in different units, normalization becomes crucial to ensure fair weighting of each criterion.

Furthermore, you correctly highlighted that there are more than just Min-Max scaling and Robust Scaling methods available to handle outliers in the data. The choice of normalization technique should indeed be carefully considered based on the data characteristics and the algorithms being used.

I understand your concern about the widespread preference for Min-Max scaling and Z-Score normalization, even though other methods may yield different and potentially better results in certain situations. It is essential for practitioners and researchers to be aware of the available alternatives and to explore different normalization strategies when dealing with diverse datasets and applications.

In conclusion, I completely agree with your viewpoint that different normalization techniques can have varying impacts on machine learning results, particularly in decision-making scenarios. As the field of AI/ML progresses, it is essential to foster a more critical and thoughtful approach to normalization, taking into account the unique characteristics of the data and the goals of the specific machine learning task.