TBT: A Deep Dive into Classification Algorithms with TensorFlow

TBT: A Deep Dive into Classification Algorithms with TensorFlow

Sometimes, looking back at past work gives us a renewed perspective on technology and its evolution. In this Throwback Tech post, I revisit my research on Deep Learning for classification algorithms using the TensorFlow library on 2018.

📄 Download the full PDF

The Experiment

This project involved testing Neural Networks with TensorFlow on a dataset of 6 million logs from the OpenCampus platform, based on OpenEdX at UTPL. The goal was to classify student actions and course approval status using a semi-supervised approach.

Data Processing Challenges

To handle such a massive dataset, I leveraged Pandas for preprocessing, ensuring clean and structured data input for the deep learning models. The Convolutional Neural Network (CNN) evaluation was performed on 1 million logs, revealing key insights into student behavior and course completion rates.

Findings & Insights

  • Two classification models were explored:
    • Event-based classification: Categorizing interactions between students and courses.
    • Approval-based classification: Predicting student success rates per course.
  • The data suggested that a majority of students struggled to pass their courses, indicating areas for potential educational improvements.

The KDD Methodology Adaptation

To enhance our approach, we adapted the KDD (Knowledge Discovery in Databases) methodology, aligning it with modern Deep Learning techniques. This hybrid methodology allowed for a more structured and effective way of extracting valuable patterns from student behavior logs.

Reflections

Looking back, this research was a pivotal step in my journey into AI and machine learning. Today, Deep Learning frameworks have evolved significantly, but the core principles remain relevant. Revisiting past experiments not only sparks nostalgia but also serves as a benchmark for future advancements.

🚀 Curious about the full study? Check it out here.

Sometimes, the best way to move forward is to look back. #TechThrowback #DeepLearning #TBT