LogLumos: Illuminating Anomalies in the Dark

https://www.kaggle.com/competitions/convolve-epoch1

- circa Dec 2022

Introduction

Welcome to the LogLumos GitHub repo! Here, you'll find the culmination of our last-minute creative sprint for the Convolve ML/AI Hackathon organized by 6 IITs in collaboration with Cisco.

Code Structure

The code structure is simple enough.

• Data Loading & Preprocessing: Initially, our code just loads the data from the kaggle website, and preprocesses it to a more suitable format. Following that we just search for a few more insights and drop a few values.

• Label Encoding: Label encoded various columns, namely, 'Details', 'ErrorType', 'ErrorLoc', 'Code0', 'Code1', 'Code2', 'Code3',and 'Code4'.

• Model Fitting: Fitted the data to a Gaussian Naive Bayes Model.

• Submission: Repeated the above steps for the evaluation data, prepared the outputs for submission as a .csv file.

Sad Story Time

The Convolve Hackathon, a cool initiative, brought together the formidable minds of data science enthusiasts for a challenging Problem Statement. The challenge was set to explore the realm of anomaly detection using log data. Although we initially didn't have grand plans for this competition due to ongoing projects and prior engagements, the allure of testing our mettle against the best motivated us to dive in with enthusiasm.

In the final hours before the submission deadline, we embarked on a coding frenzy to tackle the log anomaly detection problem. Leveraging our collective expertise with a mix of excitement and loads of sleep-deprivation, we channeled our energy into crafting what we thought was a decent enough submission. Unfortunately, due to a small issue with blank spaces in column names, we couldn't put forward our submission to be tested against those of our peers. While the finish line was tantalizingly close, this unforeseen twist, one can say, taught us the importance of meticulous attention to detail, even during the most exhilarating sprints. Or did it?

Made with ❤️ by Team BlueLock

Questions? Feel free to reach out to any of us at

• Adarsh Lal
• Gourav Kumar Sahoo
• Ashutosh Panda (<--- This me. Just in case some bloke couldn't figure it out.)