Topic 7 – Human Vital Signs (Kaggle)

Dataset

• Source: Human Vital Sign Dataset – Kaggle
• Link: https://www.kaggle.com/datasets/nasirayub2/human-vital-sign-dataset

Download Instructions

1. Open the dataset link above.
2. Click "Download".
3. Extract to data/vital_signs/.
4. Use the main CSV, e.g. HumanVitalSigns.csv.

Data Loading

import pandas as pd
df = pd.read_csv("data/vital_signs/HumanVitalSigns.csv")  # adjust
df["Time"] = pd.to_datetime(df["Time"])  # adapt time column name
df = df.set_index("Time").sort_index()

Implementation Steps

1. Data Exploration

• Load vital signs dataset
• Identify available signals (Heart Rate, Blood Pressure, SpO2, etc.)
• Inspect data frequency and time range
• Check for missing values and data quality
• Understand measurement units and normal ranges

2. Exploratory Data Analysis (EDA)

• Plot each vital sign over time
• Identify patterns (resting vs active periods, circadian rhythms)
• Calculate basic statistics (mean, std, min, max)
• Analyze relationships between different vital signs
• Perform time series decomposition

3. Signal Preprocessing

• Handle missing values (interpolation, forward fill)
• Detect and handle outliers (physiological limits)
• Smooth noisy signals if needed (moving average, median filter)
• Resample to consistent frequency if needed

4. Stationarity Analysis

• Test each signal for stationarity
• Vital signs may have trends (e.g., HR during exercise)
• Apply differencing or detrending if needed
• Consider segmenting by activity state

5. Model Building

• Univariate Models:
• ARIMA/SARIMA for each vital sign
• Consider seasonality (circadian rhythms)
• Multivariate Models (optional):
• VAR models to capture relationships between vital signs
• Analyze how one sign affects another

6. Model Evaluation

• Split data temporally
• Generate forecasts for each vital sign
• Calculate accuracy metrics
• Visualize forecasts with actual values
• Analyze forecast errors

7. Clinical Interpretation

• Interpret results in clinical context
• Identify normal vs abnormal patterns
• Discuss implications for monitoring
• Compare with clinical knowledge

Expected Deliverables

1. EDA Report:

• Time series plots for each vital sign
• Statistical summaries
• Relationship analysis between signs
• Decomposition plots

1. Model Results:

• Model parameters for each signal
• Forecast accuracy metrics
• Forecast plots
• Clinical interpretation

1. Code:

• Complete Python notebook
• Preprocessing functions
• Visualization utilities

Tips

• Vital signs have physiological limits - use domain knowledge for validation
• Consider activity states (rest, exercise, sleep) in analysis
• Circadian rhythms may create daily seasonality
• Handle outliers carefully - they may be real events (e.g., spikes during activity)
• Use appropriate sampling frequency (too high may be noisy, too low loses information)
• Multivariate models can capture physiological relationships (e.g., HR and BP)
• Consider patient-specific models vs population models
• Document any preprocessing decisions and their rationale