Data Analysis Archives

February 18 2025

Love At First Swipe: A brief look into credit cards

Today’s project will be about one of my favourite things: credit cards. I think of them as a debit card with an extra step: instead of the money coming from your checking account, it comes from the bank, and you just pay it off using your checking account. Of course, this assumes account holders have enough self-control as to not spiral out of control and accrue massive debts by overspending. But, if you know how to use them responsibly, they can be one of the best things ever. From perks, rewards to securing bigger lines of credit in the future, I think they’re worthwhile if one can be responsible.

Methods

I created a decision tree model which predicts a person’ s credit utilization ratio is high or low based on different factors such as their credit limit and revolving balance of their account. Most credit bureaus state that a person should not use more than 30% of their available credit, so that’s what I used to create the High-Low split of the model. I then analyzed this model by creating a confusion matrix of predications and assessing the model’s accuracy as well as any limitations. I also did a financial evaluation using credit card interest rates and an amortization schedule for paying off credit card debt to see just how long it takes.

Analysis

The graph shows that there is an overall negative relationship between credit limits and average utilization ratios. This is unsurprising a fixed amount spent with a lower limit will have a higher ratio compared to a higher limit.

The decision tree created had an accuracy of 96.8% when classifying new data, which is quite high. The corresponding confusion matrix yielded a false positive rate of 3.58% and a false negative rate of 2.45%, both very low when compared to the sensitivity and specificity, both above 95%. This implies that the model is a very good predictor of whether someone has a high or low credit usage based on their credit limit and total revolving balance at any given point in time. However, it is important to note that this model would need further testing on other datasets to confirm its applicability as it may conform to this data, but not to others.

Based on the amortization, we see that it takes 35 months, or approximately 3 years, to pay off a $5000 debt assuming consistent monthly payments of $200 are made at the end of each month. This is a quite some time which assumes that a person can put aside that money for this specific purpose. We also see that as more of the principal is repaid, the lower the interest accrued becomes, suggesting a negative association.

If the monthly payments were higher or payments were made on a different time basis, such as fortnightly, this would take less time as less interest is accrued, so the balance decreases to zero much faster. But if the monthly payments are less of if you miss a month or two, this will set you back even further as more interest accrues.

Conclusion

Credit cards can be very useful if used responsibly. If not, they can accrue massive debts that will take many months, even years, to pay off.

January 1 2025

All that Glitters: Analyzing Precious Metals

Ash Data Analysis 0

I’ll be doing a brief analysis of gold and silver investments using sourced data.

Introduction

Gold and silver have been prized by people for thousands of years. One of the oldest forms of currency, they were associated with high value due to their rarity and lustrous appearance. Today, gold and silver have many industrial applications in the production of jewellery, coins and even electronics due to their high electrical conductivity and resistance to corrosion. Also, they are still used as a form of currency but not in the generic sense. Because of their scarcity and demand in industry, they are treated as commodities and can be bought in the physical forms such as coins, bars, bullion or assets such as stocks in mining companies and exchange-traded funds. Investors often buy silver and gold as a hedge against inflation as they appreciate over time. But this raises the question: Are these precious metals worth the hype?

Methodology

Data was sourced from Kaggle, an online platform with free datasets and used this as a basis for my analysis. I then used R coding to clean and manipulate this data to create visualizations and calculate the overall rate of return for both gold and silver separately. I then used this data to form a linear optimization problem with constraints to maximize an initial investment and calculate the amount of silver and gold needed to achieve it. Lastly, I used some financial mathematical techniques to determine how much gold and silver you need to buy per year to accumulate to a fixed amount a decade later.

Graphs

Techniques

You can view the techniques I used by clicking the relevant word below:

R code

Optimization

Payment Valuation

Analysis

Based on the graphs of prices and return rates, we see that both gold and silver had an overall positive increase over the last few decades. There are noticeable spikes for the years 1980 and 2008, with both being attributed to extreme geopolitical and market events (for example, the collapse of the Bretton Woods system which affected the volatility of the U.S. dollar in the 1970s coupled with rampant inflation in the early 1980s and the more recent 2008 Financial Crisis). Nonetheless, the average return rate for silver was 3.96% and 6.57% for gold, which are quite impressive.

Observing the optimization problem, we see that $60000 should be invested in gold and $40000 should be invested in silver to reap the maximum return on investment, found to be approximately $5526 per year, which is not insignificant by any means (You can buy yourself a new wardrobe or take a nice vacation with that!)

Lastly, valuing some payment streams using silver and gold prices showed that you need to buy less gold than silver to get the same value, which makes sense given the higher return rate. But that consistently investing in both silver and gold would lead to hefty returns in the future.

Limitations

1. Quite a few assumptions were used for these models, so actual quantities will vary depending on future expected and unexpected global events.
2. My analysis heavily relied on the rates I found to be accurate, but more data would have painted a better picture.

Conclusion

Based on all the data and analysis above, gold and silver seem to be very good investments, both in the short and long term. Therefore any investor, whether new or veteran, should have these in their portfolio (This is NOT professional advice, so please weigh all risks appropriately if you are serious about investing).