import pandas as pd import matplotlib.pyplot as plt import plotly.graph_objects as go import numpy as np import seaborn as sns from scipy.stats.mstats import winsorize import streamlit as st df = pd.read_csv("Life Expectancy Data.csv") st.title('Analyzing The World :earth_africa:') st.write('**Below data is edited for better analysis and has 2900 rows.It gives life expectancy info for every country between the years 2000-2015. We will get so see the development or regession for each coutry and the world average.**') df.rename(columns = {" BMI " :"BMI", "Life expectancy ": "Life_expectancy", "Adult Mortality":"Adult_mortality", "infant deaths":"Infant_deaths", "percentage expenditure":"Percentage_expenditure", "Hepatitis B":"HepatitisB", "Measles ":"Measles", "under-five deaths ": "Under_five_deaths", "Total expenditure":"Total_expenditure", "Diphtheria ": "Diphtheria", " thinness 1-19 years":"Thinness_1-19_years", " thinness 5-9 years":"Thinness_5-9_years", " HIV/AIDS":"HIV/AIDS", "Income composition of resources":"Income_composition_of_resources"}, inplace = True) df.groupby('Country').apply(lambda group: group.interpolate(method= 'linear')) imputed_data = [] for year in list(df.Year.unique()): year_data = df[df.Year == year].copy() for col in list(year_data.columns)[4:]: year_data[col] = year_data[col].fillna(year_data[col].dropna().median()).copy() imputed_data.append(year_data) df = pd.concat(imputed_data).copy() df['Life_expectancy'].fillna(df['Life_expectancy'].mean(), inplace=True) df.reset_index(inplace=True) df = df.drop('index', axis=1) st.dataframe(df) st.write('**For a better Analysis, we should also remove outliers. Lets see them first.**') col_dict = {'Life_expectancy':1,'Adult_mortality':2,'Infant_deaths':3,'Alcohol':4,'Percentage_expenditure':5,'HepatitisB':6,'Measles':7,'BMI':8,'Under_five_deaths':9,'Polio':10,'Total_expenditure':11,'Diphtheria':12,'HIV/AIDS':13,'GDP':14,'Population':15,'Thinness_1-19_years':16,'Thinness_5-9_years':17,'Income_composition_of_resources':18,'Schooling':19} fig = plt.figure(figsize=(20,30)) for variable, i in col_dict.items(): plt.subplot(5, 4, i) plt.boxplot(df[variable]) plt.title(variable) plt.grid(True) st.pyplot(fig) st.write(""" We'll remove outliers in Infant_Deaths, Measles, and Under_five_deaths columns since values beyond 1000 are unrealistic. Similarly, we'll address extreme values in Expenditure_Percentage, GDP, and Population columns by taking logarithmic values. BMI values above 40 indicate extreme obesity, and some countries have averages around 60, which is not possible. Therefore, we'll remove the entire BMI column. For other columns with outliers, we'll apply winsorization for data normalization. """) # Remove outliers and log transform df = df[df[['Infant_deaths', 'Measles', 'Under_five_deaths']].lt(1001).all(axis=1)] df.drop('BMI', axis=1, inplace=True) df[['Percentage_expenditure', 'Population', 'GDP']].apply(np.log) df.replace([np.inf, -np.inf], 0, inplace=True) # Winsorization cols_to_winsorize = ['Life_expectancy', 'Adult_mortality', 'Alcohol', 'HepatitisB', 'Polio', 'Total_expenditure', 'Diphtheria', 'HIV/AIDS', 'Thinness_1-19_years', 'Thinness_5-9_years', 'Income_composition_of_resources', 'Schooling'] winz_cols = [col for col in cols_to_winsorize] df[winz_cols] = df[cols_to_winsorize].apply(lambda x: winsorize(x, limits=((0.05, 0) if x.name == 'Life_expectancy' else (0, 0.04) if x.name == 'Adult_mortality' else (0.0, 0.01) if x.name == 'Alcohol' else (0.20, 0.0) if x.name == 'HepatitisB' else (0.20, 0.0) if x.name == 'Polio' else (0.0, 0.02) if x.name == 'Total_expenditure' else (0.11, 0.0) if x.name == 'Diphtheria' else (0.0, 0.21) if x.name == 'HIV/AIDS' else (0.0, 0.04) if x.name == 'Thinness_1-19_years' else (0.0, 0.04) if x.name == 'Thinness_5-9_years' else (0.05, 0.0) if x.name == 'Income_composition_of_resources' else (0.03, 0.01)), axis=0)) # Plot boxplots for winsorized variables fig, axs = plt.subplots(3, 6, figsize=(20, 20)) cols_to_plot = winz_cols + ['Measles', 'Infant_deaths', 'Under_five_deaths', 'GDP', 'Population', 'Percentage_expenditure'] for ax, col in zip(axs.flat, cols_to_plot): sns.boxplot(y=df[col], ax=ax, color="green") ax.set_title(col) ax.set_ylabel('') ax.grid(True) plt.tight_layout() st.pyplot(fig) st.write('**Analysis**') fig = plt.figure(figsize=(20, 20)) for i, variable in enumerate(cols_to_plot, start=1): plt.subplot(6, 6, i) plt.hist(df[variable]) plt.title(variable) plt.ylabel('') plt.grid(True) st.pyplot(fig) # Plot correlation heatmap life_exp = cols_to_plot + ['Year'] plt.figure(figsize=(15, 10)) corr_matrix = df[life_exp].corr().values st.pyplot(sns.heatmap(df[life_exp].corr(), annot=True, linewidths=4).figure) # Get correlations flattened_corr = corr_matrix.flatten() sorted_corr_indices = np.argsort(flattened_corr) top_25_pos_corr_indices = sorted_corr_indices[-70:-1] top_25_pos_corr_indices = top_25_pos_corr_indices[::-1] top_25_neg_corr_indices = sorted_corr_indices[:50] # Create DataFrames for positive and negative correlations corr_columns = df[life_exp].columns corr_df = pd.DataFrame(columns=['1', '2', 'Correlation']) neg_corr_df = pd.DataFrame(columns=['1', '2', 'Correlation']) # Populate DataFrames for idx in top_25_pos_corr_indices: row, col = np.unravel_index(idx, corr_matrix.shape) if row != col: corr_df = pd.concat([corr_df, pd.DataFrame({'1': [corr_columns[row]], '2': [corr_columns[col]], 'Correlation': [corr_matrix[row, col]]})]) for idx in top_25_neg_corr_indices: row, col = np.unravel_index(idx, corr_matrix.shape) if row != col: neg_corr_df = pd.concat([neg_corr_df, pd.DataFrame({'1': [corr_columns[row]], '2': [corr_columns[col]], 'Correlation': [corr_matrix[row, col]]})]) # Drop duplicates from both DataFrames corr_df.drop_duplicates(subset=['Correlation'], inplace=True) neg_corr_df.drop_duplicates(subset=['Correlation'], inplace=True) # Display the top correlations st.write("Top 25 Positive Correlations:") st.dataframe(corr_df) st.write("Top 25 Negative Correlations:") st.dataframe(neg_corr_df) st.write(""" Key insights from the correlation analysis: - Adult mortality exhibits a negative correlation with schooling and income composition, while it positively correlates with HIV/AIDS. - Infant deaths and under-five deaths are strongly positively correlated. - Schooling and alcohol consumption display a positive relationship. - Percentage expenditure shows positive correlations with schooling, income composition, GDP, and life expectancy. - Hepatitis B is strongly positively correlated with polio and diphtheria. - Polio and diphtheria show strong positive correlations with each other and with life expectancy. - Life expectancy is positively correlated with schooling, income composition, GDP, diphtheria, polio, and percentage expenditure. Conversely, it is negatively correlated with adult mortality, thinness in both age ranges, HIV/AIDS, under-five deaths, and infant deaths. """) # GRAPHS df['Status'] = df['Status'].map({'Developed': 1, 'Developing': 0}) def plot_by_country_development(data, value_column, value_title): value_year = data.groupby(['Year', 'Status'])[value_column].mean().unstack('Status').fillna(0) value_year.columns = ['Developing', 'Developed'] fig = go.Figure() fig.add_trace(go.Scatter(x=value_year.index, y=value_year['Developing'], mode='lines', name='Developing', marker_color='#f075c2')) fig.add_trace(go.Scatter(x=value_year.index, y=value_year['Developed'], mode='lines', name='Developed', marker_color='#28d2c2')) fig.update_layout(height=500, xaxis_title="Years", yaxis_title=value_title, title_text=f'{value_title} Average of Countries Over The Years', template="plotly_dark") return fig st.plotly_chart(plot_by_country_development(df, 'Life_expectancy', 'Life Expectancy')) st.plotly_chart(plot_by_country_development(df, 'Schooling', 'Schooling Level')) st.plotly_chart(plot_by_country_development(df, 'Income_composition_of_resources', 'Income Composition of Resources')) st.write("### Population Analysis") fig_hiv = plot_by_country_development(df, 'Thinness_5-9_years', '5-9 years old population') fig_diptheria = plot_by_country_development(df, 'Thinness_1-19_years', '1-19 years old population') fig_polio = plot_by_country_development(df, 'Adult_mortality', ' Adult Mortality') fig_hepatitisB = plot_by_country_development(df, 'Infant_deaths', 'Infant Deaths') height = 400 width = 400 fig_hiv.update_layout(height=height,width=width) fig_diptheria.update_layout(height=height, width=width) fig_polio.update_layout(height=height, width=width) fig_hepatitisB.update_layout(height=height, width=width) col1, col2 = st.columns(2) with col1: st.plotly_chart(fig_hiv) with col2: st.plotly_chart(fig_diptheria) with col1: st.plotly_chart(fig_polio) with col2: st.plotly_chart(fig_hepatitisB) st.write("### Diseases Analysis") fig_hiv = plot_by_country_development(df, 'HIV/AIDS', 'HIV/AIDS') fig_diptheria = plot_by_country_development(df, 'Diphtheria', 'Diphtheria') fig_polio = plot_by_country_development(df, 'Polio', 'Polio') fig_hepatitisB = plot_by_country_development(df, 'HepatitisB', 'HepatitisB') height = 400 width = 400 fig_hiv.update_layout(height=height,width=width) fig_diptheria.update_layout(height=height, width=width) fig_polio.update_layout(height=height, width=width) fig_hepatitisB.update_layout(height=height, width=width) col1, col2 = st.columns(2) with col1: st.plotly_chart(fig_hiv) with col2: st.plotly_chart(fig_diptheria) with col1: st.plotly_chart(fig_polio) with col2: st.plotly_chart(fig_hepatitisB)