DNA-Classification-Challenge-With-Kernel-Methods

This solution ranked 1st in the inclass competition on Kaggle out of 43 teams. The data challenge is a project taken to climax the kernel methods in machine learning course at AMMI-2020, aimed at the implementation of machine learning algorithms to gain understanding and further adapt them to structural data (DNA sequence data). The report summaries the approach adopted by Aissatou Ndoye and I to the challenge which was hosted on Kaggle with the goal of predicting whether a DNA sequence region is binding site to a specific transcription factor or not. Our best result ranked 1st on the private leader board with a score of 71.20%. It is worth noting that increasing the value of lambda from 0.80 through to 1 improves the private score to 71.40%. Moreso, it was later identified that a mismatch of 2 in both first and second kernels & 3 in third produces a public score of 71.60 and private score of 72.60.