Machine Learning

Team Members :SAGUER Quentin/GUESSOUS Samy CARRE Pablo/PONTHIEU Gabriel/TALLARON Matéo

Machine Learning

Dataset Overview

• Source : Cyber Threat Detection document

• This dataset contains 1430 rows and 23 columns

• Problem statement :
• Goal : Classify network activities as either malicious or mild

• Target : Label column where 1 = malicious and mild = 0

Exploring Dataset

Data Cleaning

• Tasks performed :

• Removed irrelevant columns

• Verified there were no missing values in the dataset

• Checked for duplicates : "No duplicate rows found"

Splitting features and target

Data Visualization

• Features : All relevant columns except Label

• Target : The Label column

• Charts to include :

• Bar chart : Distribution of Label values

• Histogram : Distribution of Packet_Length or another numeric features

• Heatmap : Correlation between features (use seaborn or similar)

Splitting Data

• Process :

• Split dataset into training (80%) and testing (20%)

• Use the python code above

• Why ?

• To ensure model is tested on unseen data

Training Models

• Models used :

• Logistic Regression

• Random Forest Classifier

• Support Vector Machine (SVM)

• Process :

• Train each model using the training data

• Use the python code here shown ->

Models Evaluation

• Key Insight :

• Class imbalance is a challenge but can be managed with Random Forest

• Key features such as Packet_Length and Bytes_Sent are critical for classification

• Visualizing the data helped in understanding distributions and feature importance, leading to a better model

Conclusion

• Best Model :

• Random Forest showed the best overall performance for this classification task

Any questions

Thank you for listening