This link talk about:

Artificial neural networks for small dataset analysis

Article Artificial neural networks for small dataset analysis

Hello Mahmuh,

A very good question. First of all, what is the type of your data? is it tabular or not? I mean is it images or just data. And if it is data, what kind of data? Numerical or categorical data? secondly, the NN should be able to generalize to unseen data, that is the whole idea of machine learning, to be able to predict unseen new data. But why NN? you should try different algorithms as well.

Hello Mr. Ibrahem,

I have a cross section set of five country's which is consisting of numerical data and variables such as economic cost of pollucition, greenhouse gas measurement values, etc.

Assuming these variables (or observations) ​​are country specific, I wondered if the estimation made for another country would be valid. I thought that artificial networks would be more suitable,

Since I'm new to machine learning techniques, I am open to suggestions of different algorithms.

Thank you so much for response.

Best regards

Hello Mahmuh Khaddi

please check the following links:

https://knoema.com/search query=pollution+dataset+argentina&pageIndex=&scope=&term=&correct=&source=Header

Deaths due to air pollution/country

https://ourworldindata.org/air-pollution

Regarding using data from other countries to predict the impact in Argentina, the estimates generated by such ML models may not be very useful in the context of Argentina as different countries have different impacts of pollution.

Also, the predictive power of ML models is as good as the data provided. try to gather data specific to Argentina from different sources, that will give more specific estimates.

Matlab

Dear Mohammad Khalid Pandit

Thank you for data sets. You made a point. I just wondered if there could be unit (country in my case) effects and can't get accurate values.

But actually I want to predict/estimate cost of pullution. Becasuse other five country's have such variables but one country.

Since the cost factors are the same for all countries, I thought I would get effective results when I made a cost estimation.

But my biggest drawback was the country effect. For this reason, I wondered if I could make cost estimates and get effective and generalizable results with artificial neural networks.

Thank you for response.

Best regards.

Yes you can..if and only if the circumstances of both countries are similar then you can train a neural network on a dataset from Brazil and use it to predict on a data set from Argentina... You can use remote data from Argentina and predict the future and compare with actual data from the recent past to check if your model is doing well... Use metrics like R-squared to check the validity of your model.

This link talk about:

Artificial neural networks for small dataset analysis

Article Artificial neural networks for small dataset analysis

Mahmuh Khaddi your question is more statistical in nature than a question on neural networks. Models trained for one country will probably be of little use for other contries.

Generally speaking, any model can be used to predict new observations, provided you have the same independent variable set.