How izangdev/prediccion-abandono-ml works
Main Components
• Data Sources: Represent the raw data files (AbandonoEmpleados.csv, abandono.xlsx) that feed into the analysis. These are the foundational inputs.
• Data Analysis & EDA Component: Represented by the Notebook_Abandono.ipynb. This component uses libraries such as pandas, numpy, and matplotlib to clean the data, handle nulls, categorize numeric and categorical features, and extract key insights.
• Machine Learning Model Component: Highlight the ML pipeline that transforms data (using ScikitLearn) and builds a decision tree model, achieving approximately 70% accuracy. Here, individual steps like data transformation and model training (including the integration of the results into the dataset) should be delineated.
• Visualization/Dashboard Component: Show Tableau as an external service used to build the final dashboard. The assets folder (e.g., Dashboard.png) supports this component by providing static assets, and the README includes a link to the live dashboard.
Relationships and Data Flow
• From EDA, show the process flow into the Machine Learning Model component where data is transformed (e.g., converting variables to numerical) and the decision tree model is trained.
• After model training, indicate that results are merged back with the original dataset.
• Finally, represent how the processed data and results are utilized by the Tableau dashboard for visualization.
Architectural Patterns and Design Principles
• Modularity: Emphasize separable modules (data input, analysis, model training, visualization) to highlight the distinct roles and responsibilities.
• Separation of Concerns: Demonstrate how the computational logic (in the Jupyter Notebook) is separated from the visualization layer (Tableau), making the system easier to manage and adapt.
Relevant Technologies, Frameworks, and Libraries
• Data Handling: Python libraries (pandas, numpy, matplotlib) used in the Notebook for initial analysis.
• Machine Learning: ScikitLearn for data transformation and decision tree classification.
• Visualization: Tableau (and associated public dashboard link) for interactive visualization.
• Data Files: CSV and XLSX formats as sources; emphasize that these are ingested and pre-processed.