How hkkiat/job-recommendation-system works
Main Components
• Data Sources: Represent the raw CSV survey data in the “Data” folder and their copies under “streamlit/Data.” These are the input feeds for both model training and on-the-fly data visualisations.
• Model Training & Storage: Include a component for the model training process from “Code for model generation.ipynb” (along with its associated checkpoints) that generates artifacts like “logreg_model.joblib” and “Oh_encoder.joblib.” This process encapsulates the preprocessing, feature selection, and model tuning for the recommendation engine.
• Helper/Library Module: Represent the “streamlit/helper” folder as a set of utility components that provide common functions (data processing, Google Colab integrations, etc.) supporting both the EDA and recommendation functionalities.
• EDA Frontend Pages: Illustrate the multipage structure under “streamlit/pages” (e.g., “Data Processing.py,” “Milestone 1 Statistic.py,” “Milestone 1 Trend Over Time.py”) that offer visualizations and dynamic analysis of trends over time and statistical insights.
• Recommendation Engine UI: Use the “Milestone 2 Job Recommendation.py” page as the user interface for receiving applicant profiles, passing them to the pre-trained model, and displaying the recommended job roles.
• Streamlit Application Container: Overall, frame the “streamlit” folder as your integrated web application that ties together the UI, helper utilities, and pre-trained model with the necessary configurations (as listed in “requirement.txt”).
Relationships & Interactions
• Connections between the Helper Modules and both the EDA pages and Recommendation UI to indicate shared utilities and preprocessing logic.
• Streamlit UI to the underlying components (EDA modules and recommendation functionality), showing the flow of user interactions into data processing and model inference.
Architectural Patterns & Technologies
• This design separates concerns between model training (offline, using Jupyter Notebooks) and model serving (online in the Streamlit app).
• The multipage design of Streamlit is a key design principle—each page handles a specific function, whether data visualization or recommendation processing.
• Use separate submodules to encapsulate data preprocessing (common_helper, data_helper, and data_google_colab) reflecting a modular design that is easy to extend.
• That you’re using pre-trained machine learning models built with scikit-learn (as indicated by “joblib” files) and the Streamlit Python framework for the frontend.
• That the saved models (joblib files) and helper modules are clearly shown as shared resources for the recommendation system.
• Finally, include configuration details if needed (reference to “requirement.txt” for package dependencies).