Skip to content

LLM Keras NLP project utilized confusion matrix to get the insight from the text mining and get a desired outcome to check the texts being generated by the AI or human.

Notifications You must be signed in to change notification settings

tahiyar7/LLM-KerasNLP

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 

Repository files navigation

LLM-KerasNLP

Hi this is Tahiya. This project has used two diffrenet sets of dataset to understand and create insights results to know how much text being generated by AI and the students. I have seen a decent result after F1 Score and consfusion matrix. :fleur_de_lis:

Workflow

  • Install Package 🧵
  • Import Library ♣️
  • Load & Explore: 🪁 Vizualization 🖼️
  • Add new dataset: 🧩 Visualization 🖼️
  • Prepare the data 🪡
  • LLM Model: Confusion Matrix 🥽

Result

The F1 score is a measure of a test's accuracy and is especially useful when dealing with imbalanced datasets. An F1 score of 0.91 indicates a high level of accuracy, meaning that the test or model is performing well in terms of both precision and recall. To better understand this, let's break down the meaning of an F1 score and the components of a confusion matrix. 🎧

Confusion Matrix Components A confusion matrix for a binary classification problem is structured as follows:

image

True Positives (TP): The number of correct positive predictions. False Positives (FP): The number of incorrect positive predictions. False Negatives (FN): The number of incorrect negative predictions. True Negatives (TN): The number of correct negative predictions. Picture taken from https://www.v7labs.com/blog/confusion-matrix-guide Ideas taken from https://f-a.nz/dev/develop-your-own-llm-like-chatgpt-with-tensorflow-and-keras/

About

LLM Keras NLP project utilized confusion matrix to get the insight from the text mining and get a desired outcome to check the texts being generated by the AI or human.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published