The purpose of this project is to follow the guidelines of the CRIP_DM to, through data mining tools, answer the following question: “Can we predict the euro exchange rate for the next day using the previous rates?”. The Foreign_Exchange_Rates.csv data set was used in this analysis.
The .csv file contains a 19-year historical series of the euro against the dollar. The static models used were: ARIMA (AutoRegressive Integrated Moving Average) and RNN (Recurrent Neural Network). The following Python libraries were used: Pandas, Matplotlib, Numpy and Seaborn.
Data analysis and the search for patterns in them are essential activities in the financial market. Investors always try to have all the information possible before making their investment decision. Foreign exchange investments are no different and the search for patterns in a historical exchange rate series is essential to predict the future exchange rate. In answering the question, we will be developing a tool that can assist investors in their analysis of the foreign exchange market.
Following the CRISP-DM guidelines, we start the project with a Business Understanding. The first stage was to understand what you want to accomplish from a business perspective.
To answer the question proposed for this project, it is necessary to carry out an exploratory and comparative statistical analysis of the Data Set. Several data mining tools will be used so that we can verify, explore and analyse the data to be used in the model. The model built will be able to use the available data and inform with a certain level of accuracy the forecast of the exchange rate for a given country.
Care must be taken in evaluating the results and in the model used so that they are statistically significant, ensuring that the conclusions are robust enough.
Steps: Data Understanding, Data Preparation, Modelling and Evaluate Results.
- CRISP-DM methodology;
- Exchange rate prediction (next day);
- Analysis of historical data;
Technical Features:
- Jupyter Notebook.
- ARIMA - AutoRegressive Integrated Moving Average and;
- RNN - Recurrent Neural Network;
- Python libraries: Pandas, Matplotlib, Numpy and Seaborn.
This project was part of the Data Mining class of my Master of Applied Software Development program. The premises and the .csv file were defined by the professor. All the code is my authorship as well as all the project development in Jupyter Notebook.
This project is finished because I understand that it fulfilled all the objectives, both in the use of the CRISP-DM methodology and in the data analysis.