app.py

import pandas as pd
import altair as alt
import streamlit as st 

# Load the data 
child_mortality = pd.read_csv("https://huggingface.co/spaces/jiyachachan/fp2/resolve/main/child_mortality_0_5_year_olds_dying_per_1000_born.csv")  # Format: Country, Year, Value
gdp_per_capita = pd.read_csv("https://huggingface.co/spaces/jiyachachan/fp2/resolve/main/gdp_pcap.csv")    # Format: Country, Year, Value

# Melt datasets to tidy format
child_mortality = child_mortality.melt(id_vars=["country"], var_name="year", value_name="child_mortality")
gdp_per_capita = gdp_per_capita.melt(id_vars=["country"], var_name="year", value_name="gdp_per_capita")

# Merge the datasets
merged_data = pd.merge(child_mortality, gdp_per_capita, on=["country", "year"])
merged_data["year"] = merged_data["year"].astype(int)  # Ensure 'year' is an integer

# Convert gdp_per_capita and child_mortality to numeric
merged_data["gdp_per_capita"] = pd.to_numeric(merged_data["gdp_per_capita"], errors="coerce")
merged_data["child_mortality"] = pd.to_numeric(merged_data["child_mortality"], errors="coerce")

# Drop rows with missing or invalid data
merged_data = merged_data.dropna(subset=["gdp_per_capita", "child_mortality"])

# Streamlit app
st.title("Interactive Visualization: GDP vs. Child Mortality")

st.text(" ")

st.text("The dataset represents global development indicators related to child mortality and GDP per capita for multiple countries over several years. Each row corresponds to a unique country-year combination, with the key fields being country (categorical, representing the country name), year (integer, indicating the year of data collection), child_mortality (numeric, showing the number of children under five dying per 1,000 live births), and gdp_per_capita (numeric, representing GDP per capita in constant 2017 international dollars). The dataset spans a wide range of years and countries, making it suitable for temporal and regional analyses. Missing values are present in some fields, particularly for earlier years or less-developed countries, and were handled during the data cleaning process. The values in child_mortality range from 2.24 to 756.0, while gdp_per_capita spans from $354.00 to $10,000.00, reflecting significant disparities in economic and health outcomes across countries and regions.")

st.text(" ")

# Filter data for a specific year
year = st.slider("Select Year", min_value=int(merged_data["year"].min()), max_value=int(merged_data["year"].max()), value=2020)
filtered_data = merged_data[merged_data["year"] == year]

# Select number of countries to display
num_countries = st.slider("Select Number of Countries to Display", min_value=5, max_value=50, value=10, step=5)

# Get top N countries by GDP per capita
top_countries = filtered_data.nlargest(num_countries, "gdp_per_capita")

# Create scatter plot with regression line
scatter_plot = alt.Chart(top_countries).mark_circle(size=60).encode(
    x=alt.X("gdp_per_capita:Q", scale=alt.Scale(type="log"), title="GDP per Capita (Log Scale)"),
    y=alt.Y("child_mortality:Q", title="Child Mortality (per 1,000 live births)"),
    color="country:N",
    tooltip=["country", "gdp_per_capita", "child_mortality"]
).properties(
    title=f"Relationship Between GDP Per Capita and Child Mortality ({year})",
    width=800,
    height=500
)

# Add regression line
regression_line = scatter_plot.transform_regression(
    "gdp_per_capita", "child_mortality", method="linear"
).mark_line(color="red")

# Combine scatter plot and regression line
final_chart = scatter_plot + regression_line

# Display chart in Streamlit
st.altair_chart(final_chart, use_container_width=True)


st.text("To build the observatory, I began by preparing the dataset, which involved merging child mortality and GDP per capita data based on common fields: country and year. I ensured that the data was cleaned and formatted correctly, converting numerical fields like child_mortality and gdp_per_capita to numeric types and handling missing values by dropping rows with invalid entries. Once the data was ready, I created initial static visualizations using Altair to explore the relationship between GDP per capita and child mortality. Building on this foundation, I added interactivity through Streamlit, allowing users to dynamically filter the dataset by year and select the number of countries to display. To enhance the visual analysis, I overlaid a regression line on the scatter plot, which provides a clear representation of trends. The app's functionality was refined iteratively, incorporating sliders for user interaction and tooltips for exploring country-specific data points.")