How aymenmed96/predicting_sleep-dreaming_for_mammals works
Project Type & Purpose
• Type: Jupyter-based data science / ML project
• Purpose:
– Part 1: Predict total sleep time for each mammal species
– Part 2: Predict dreaming (REM) time for each mammal species
– Part 3: Correlate biological & ecological features
A. Data Sources
– Famille_Taxonomie.xlsx (taxonomy metadata)
– sleep_merged.tsv (merged features & sleep data)
B. Data Ingestion & Preprocessing
– Missing-value handling (Totalsleep-Fill-NA notebooks, Totalsleep-NA delete)
– Data merging & cleaning
C. Exploratory Data Analysis (EDA)
– Correlation analysis workbook (Correlation between Biological and Ecological attributes.ipynb)
– Visualization libraries (Matplotlib, Seaborn, etc.)
D. ML Pipelines (two parallel flows)
1. Total-Sleep Prediction Pipeline
• Notebook: Totalsleep-Prediction notebooks
• Steps: feature selection → train/test split → model training (e.g., RandomForest, SVR) → evaluation
• Dependencies: requirements-totalsleep-prediction.txt
2. Dreaming-Time Prediction Pipeline
• Notebook: Dreaming_prediction.ipynb
• Steps: same as above, different target
• Dependencies: requirements-dreaming_prediction.txt
E. Reporting & Outputs
– PDF report (Report_ML Project_BENGARAA_TOUATI_KHERARBA_COLLIN.pdf)
– Figures & tables embedded in notebooks or exported images
F. Environment & Tools
– Jupyter Notebook server
– Python runtime & virtual environments per requirements-*.txt
Relationships & Data Flows
• Data Sources → Ingestion/Preprocessing
• Preprocessing → EDA and both ML pipelines
• ML pipelines → Evaluation & Reporting
– Show that the preprocessing step feeds all three parts (total sleep, dreaming, correlation) in parallel.
Architectural Patterns & Tech Stack
– Pattern: Linear/Branching Data Pipeline (Ingestion → Processing → Modeling → Reporting)
– Libraries: pandas, numpy, scikit-learn, matplotlib, seaborn, openpyxl (for Excel), TSV reader
– Environment management: requirements files for each sub-project