Hi

Look at the CA_Markov method in IDRISI!

Also, there are several machin learning packages provided in python and R .

You should divide your data in two sections, first to predict and modeling, and second, for validation. After making satisfactory model based on your data then you be able to forecast future condition. Use the data of 2017 to validate and go on...

Good luck

you will have to use the spatial statistics tool (OLS) to run a correlation between the particulate matter and the years. if there is a correlation then you will use the model to predict.

you can only use Markov chain for landuse and not Particulat matter (P.M 2.5)

You need, first of all, to know the law of variation of your data and consequently find a model that is capable of reproducing them. On this basis, you can use multiple tools, available in GIS software, to extrapolate your data

For this you can use RCP analysis, AHP Model and Maxent too to prepare predict map .......

You can do prediction modeling using cellular automata. The temporal data can be processed using the IDRISI SELVA software.

Good luck

You can use Techniques like Land use Change Modeling (LCM), CA (Cellular Automata)-Markov process model for temporal analysis and Multi Layer Perceptron Markov Model in IDRISI Software.

You can also read these articles for Prediction mapping:

https://www.researchgate.net/publication/328744113_Markov_Chain_Modelling_of_Land_Cover_Changes_in_Jodhpur_City.

https://www.researchgate.net/publication/328383222_Prediction_of_Land_Cover_Changes_of_Jodhpur_City_Using_Cellular_Automata_Markov_Modelling_Techniques.

https://www.researchgate.net/publication/330366868_Modelling_and_Prediction_of_Land_Use_Changes_in_Jodhpur_City_using_Multi-Layer_Perceptron_Markov_Techniques

Good luck