← All projects

Python · LLMs · Logistic Regression

Comparative Analysis of LLM Performance vs Traditional ML Techniques

An empirical comparison of large language models and conventional machine-learning methods for text classification.

Year
2025
Stack
Python, LLMs, Logistic Regression

The problem

It is not always clear whether an LLM provides enough classification benefit to justify its additional complexity compared with established machine-learning models.

Approach

We evaluated large language models alongside logistic regression, support vector machines, and random forests on the same text-classification task. Each model was measured with consistent evaluation criteria so that differences in precision, recall, and F1 score could be compared directly.

Analysis

Beyond aggregate scores, the project examined incorrectly classified examples to identify where each approach performed well and where its assumptions or representation produced errors.

Outcome

Compared model performance using F1 score, precision, and recall, then analyzed the trade-offs and common causes of misclassification.

PythonLLMsLogistic RegressionSVMRandom ForestData Analysis
Esc
HomeOverviewWorkExperienceProjectsCase studiesAboutBackgroundRésuméPrint viewGitHub ↗CodeLinkedIn ↗Connect