Foreign Exchange Rates Ensemble for Classification of Trends

classification
machine_learning
python
finance

#1

I’m currently handling a project and I need an advice, please. I refer to the paper “Multi-Scale Foreign Exchange Rates Ensemble for Classification of Trends in Forex Market”. Visit link https://www.sciencedirect.com/science/article/pii/S1877050914003676.

My project:

I have hourly data on the number of transactions made by customers and the respective avg. hourly FX rate (USD/BDT). There is an inverse relationship between the number of transactions and FX rate. When the FX rate increase, the number of transactions decrease.

My objective out of this project is:

  1. To identify the up, down and sideways in the fx to be able to do the below.

  2. To classify EACH customer whether he is fx sensitive i.e. he checks the fx rate closely and transact or he does not care about fx and transact or he is neutral. For example, if most of the time, we note that he transacts when the fx drops, then we can say he is fx sensitive.

In the paper, they used a threshold of 10%. From my side, I’m thinking of calculating the threshold based on the volume of transactions. Example: If we noted that when the FX drop or increase by 5%, the % change in the volume is more significant, then I will use 5% as the threshold. Does that make sense, please?

Also, can you please help me in understanding as from point 2.2 - as from 3rd paragraph and point 2.3 in the research paper. It’s with regard to the feature and sampling rates.

Can you please advice me how do I approach this project otherwise.

Thanks,

Regards,
Nabiil.