IBM NOAA Project: Predicting the Rain: A Data Science Journey through JFK Airport Weather Patterns

 

By Matthew Ternenge Beeun

Weather forecasting has always felt like a mix of high-level physics and a bit of magic. But as a Data Scientist, I know it’s actually a mix of high-quality data and robust statistical modeling. I recently completed a deep-dive analysis into the local climatological data from John F. Kennedy International Airport (JFK). Using R Studio, I set out to answer a fundamental question for a weather forecast firm: Can we accurately predict daily precipitation using variables like temperature, wind speed, and pressure?

The Approach: The 7-Step Workflow

I followed a rigorous data science lifecycle to ensure the insights were both statistically sound and business-relevant:

  1. Data Ingestion: Sourcing historical NOAA records.

  2. Wrangling: Cleaning "Trace" values and converting raw strings into usable numbers.

  3. Exploratory Data Analysis (EDA): Using histograms and scatterplots to see the "shape" of the weather.

  4. Modeling: Building a Multiple Linear Regression engine.

  5. Prediction: Testing the engine on "unseen" historical dates.

  6. Visual Assessment: Using heatmaps to identify our "MVP" (Most Valuable Predictor) variables.

  7. Numerical Validation: Calculating error metrics like MSE and RMSE.

What the Data Told Us

The most exciting part of any analysis is the "Aha!" moment. By plotting a Correlation Heatmap, I was able to single out exactly what drives the rain at JFK.

  • Humidity is King: There is a strong positive correlation (0.41) between relative humidity and rainfall.

  • The Pressure Signal: We saw a clear negative correlation (-0.21) with station pressure. In simple terms: When the barometer drops, grab your umbrella.

  • The Model’s Accuracy: Our model achieved a Root Mean Squared Error (RMSE) of 0.22. This means that on average, our forecast is off by less than a quarter-inch of rain—a respectable margin for a linear model in a highly variable environment.

The visuals that support these assumptions are below:


The Challenge of "Zero-Inflation"

One of the key takeaways from this project was the difficulty of predicting "No Rain" days. Because most days at JFK are dry, the linear model tends to over-predict small amounts of moisture. This is a classic "zero-inflation" problem in meteorology, and it highlights why data scientists are constantly iterating on their models!

Final Thoughts

This project was a great reminder that data isn't just numbers on a spreadsheet; it’s a story about our environment. By leveraging R and Multiple Linear Regression, we can turn raw atmospheric observations into actionable insights for airport operations and public safety.

Check out my full code and the cleaned dataset on my GitHub! Github

  • Keywords: Use tags like #RStats, #DataScience, #Meteorology, and #MachineLearning.

Comments

Post a Comment

Popular posts from this blog

Google Analytics Case Study Project.

Image
   Welcome to My Data Analytics Portfolio. I am Mr Matthew Ternenge Beeun, an aspiring data analyst with a passion for transforming raw data into actionable insights. This portfolio showcases my work on the Bellabeat case study, where I applied SQL, R, and Tableau to clean, analyze, and visualize user behavior data from smart wellness devices. The goal of this project was to uncover patterns that could inform Bellabeat’s business strategy and marketing decisions. Below, you’ll find my datasets, code files, visualizations, and stakeholder report—all organized for clarity and impact. Table of Contents 1. Project Overview 2. Data Cleaning (SQL) 3. Visualizations. (Tableau) 4. Stakeholder Report (R Markdown) 5. Downloadable Files 6. External Links(Github, Kaggle) 1. Project Overview Bellabeat Case Study This portfolio project explores user behavior data from Bellabeat, a wellness tech company that manufactures smart devices for tracking fitness, activity, sleep, and health metr...
Read more

Google Data Analytics Project 2: Bike Share Case Study Project

Image
  Transforming Data into Strategy: How I Converted Cyclistic 's Casual Riders into Members. Published on:  18th October, 2025 Tags: #DataAnalytics #CaseStudy #RStudio #Tableau #Cyclistic #MarketingStrategy Introduction: The Business Problem As a Data Analyst for the fictional bike-share company Cyclistic , my core mission was simple: How do annual members and casual riders use Cyclistic bikes differently, and how can we use that insight to convert casual riders into annual members? This project moved through the entire data analysis process—from Ask to Act—culminating in clear, data-backed marketing recommendations for the Director of Marketing. 1. Ask & Prepare: Defining the Challenge The business objective was to drive membership growth. We started with 12 months of Cyclistic trip data (publicly available historical data) and prepared it for analysis. Below are the datasets that are made available for this project:   Initial Challenge: The data was inconsistent, ...
Read more

Teach your Child Coding

Image
  Why Every Child Needs a Coding Adventure: A Guide for Modern Parents ​In today’s fast-paced world, we often hear the term "digital native" applied to our children. They can swipe, tap, and navigate a smartphone before they can even tie their shoelaces. But there is a profound difference between being a passive consumer of technology and being an active creator of it. ​At Bummptech Global Concept , we believe that learning to code is about much more than just computers—it’s about giving your child the tools to build their own future. ​The New Universal Language ​Just as we learn to read and write to navigate society, coding has become the "new literacy." It is the language of the modern world. Whether a child grows up to be a software engineer, a doctor, a designer, or a businessman, the ability to understand how digital systems work will be their greatest competitive advantage. ​4 Reasons to Start the Coding Journey Early ​1. From Consumption to Creation ​M...
Read more