How abdur1r/fake-profile-detection-using-ml works
Project type and purpose
• This is a data-science / machine-learning pipeline whose goal is to detect fake profiles in online social networks. The README and directory structure show multiple modeling experiments (Neural Network, Random Forest, SVM) implemented both as Jupyter notebooks and standalone Python scripts, with input data in CSVs and output reports in HTML/PDF.
File structure
• Top-level:
• Neural Network.ipynb & .py
• Random Forest.ipynb & .py
• Support Vector Machine.ipynb & .py
• data/ (fusers.csv, users.csv)
• html/ (model report outputs)
• pdf/ (model report outputs)
• requirement/ (presumably dependency listing)
• Pattern: each algorithm is a separate module; data lives in a “data” bucket; outputs are captured in “html” and “pdf.”
README for additional insights
• Dependencies: numpy, pandas, matplotlib, scikit-learn, pybrain, sexmachine.
• Execution modes: Jupyter notebooks (recommended) or Python scripts.
• All experiment outputs are saved in HTML/PDF for easy sharing.
a. Main components
1. Data Storage Layer
– CSV files under data/
2. Data Preprocessing Module
– routines in each .py/.ipynb that load CSVs, clean, feature-engineer (e.g., using pandas, sexmachine for gender inference)
3. Model Training & Evaluation Modules
– Neural Network (pybrain)
– Random Forest (scikit-learn)
– Support Vector Machine (scikit-learn)
4. Reporting/Export Module
– generates HTML (in html/) and PDF (in pdf/) via notebook export or Python script
5. Execution Environment
– Jupyter Notebook server or Python CLI
6. Dependency Manager
– pip with requirement files
b. Relationships & Data Flow
– The Execution Environment orchestrates end-to-end flow; Dependency Manager provides libraries
c. Architectural Patterns & Design Principles
– Modular experimentation: each algorithm is isolated in its own script/notebook
– Pipeline pattern: linear flow from raw data to reports
– Separation of concerns: data ingestion, model logic, and reporting are distinct
d. Technologies/Frameworks
– pandas & numpy for data ops
– sexmachine for gender inference
– scikit-learn for Random Forest & SVM
– pybrain for neural networks
– matplotlib for visualizations
– Jupyter Notebook/IPython for interactive exploration/export
Diagramming recommendations
• Blue = Data / Storage
• Green = Compute / Models
• Orange = Reporting / Export
• Gray = Environment / Dependencies
Detail level
• Sub-components in each module (e.g., “Feature Engineering,” “Model Training,” “Model Evaluation”).
• That requirement/ (or requirements.txt) feeds into the Dependency Manager.
• That each algorithm’s notebooks/scripts import the preprocessing code.