Blog
Commodity Price Prediction (Part I)
I’ve been fascinated by the financial/trading sector since probably my mid-20s. As a part of this fascination, I wanted to use machine learning to see if we can build a model to predict commodity futures prices within a 5% error rate. The dataset that I used in this project was the World Bank Commodity Price Kaggle dataset, spanning from 1960 to 2026. It covers the monthly price of 71 commodities across 10 broad categories. The primary source of the data is from the World Bank Pink Sheet (CMO Historical Data Monthly) which spanned to end 2024; the data from Jan 2025 - Feb 2026 was sourced from the FRED API. Since then, the link to the dataset on Kaggle broke - although the data still exists on World Bank site here. The dataset I used is still available on my GitHub repo. Since we also needed to ensure that the prices were adjusted for inflation, the target column that we are using in this project is the price_index_2000_base, and not price_nominal_usd.
July 09, 2026
Housing Prices across university/non-university towns
The project covered in this post is based on UMichigan’s Intro to DS in Python Week 4 on Coursera. The repo and datasets are available on my GitHub. The purpose of the project was to compare the impact of recession on housing prices across university/non-university towns. The datasets used in this project were university_towns.txt (list of US-based universities and their towns) world_bank.csv (CSV of countries with their annual GDP from 1960 to 2015) City_Zhvi_AllHomes.csv (CSV of housing prices by month across regions and states) Hypothesis: University towns housing prices are less affected by recession compared to non-university housing prices.
October 19, 2017