Of Rising ScholarsFall 2022
Alexander will be presenting at The Symposium of Rising Scholars on Saturday, September 24th! To attend the event and see Alexander's presentation,Register here!
Bridgewater Raritan High SchoolClass of 2023Bridgewater, New Jersey
- "What are the most effective ways to use machine learning to predict systemic crises in African countries?" with mentor Nick (Sept. 4, 2022)
What are the most effective ways to use machine learning to predict systemic crises in African countries?
Started June 2, 2022
Abstract or project description
In third world countries such as the ones in Africa, systemic crises have historically devastated entire populations of people. Different factors have led to these systemic crises, including inflation, debt defaults, and other factors. The occurrence of these factors could be used with machine learning models in order to predict the occurrence of systemic crises. Predicting systemic crises is an important first step towards preventing them or minimizing their effects in these countries. In order to do this, we used a systemic crises dataset on Kaggle. We used a number of methods to clean up the data set in order to get it ready for machine learning algorithms. We dropped columns that would be unnecessary or detrimental to predictions, such as 'country' data. We then converted binary string data, which was labelled "crises" and "no_crises", to binary values to allow for numeric matrices. Next, we subtracted a minimum value from all data values in the 'year' column to minimize their weighting. Finally, we standardized two columns, 'exch_usd' and 'inflation_annual_cpi'. This was done by first capping their values at 500 and -500, then using the mean and standard deviations of the columns in order to calculate each element's z-score. Standardization was done in order to make sure their values would be weighted equally in a machine learning model. After cleaning up our data, we will train a number of machine learning models to see which one would best predict systemic crises. This included perceptron, decision trees, random forests, and other machine learning models.