app.py

import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
import joblib

@st.cache_resource
def load_dataset(path: str) -> pd.DataFrame:
    """ 
    Opens up a `.csv` file as our main dataset.
    
    Args:
        path (str): Path to the dataset to be opened
        
    Returns:
        pd.DataFrame: A pandas dataframe
    """
    
    df = pd.read_csv(path)
    
    dates = pd.to_datetime(df['Date'])
    df['Date'] = dates

    df['Year'] = df['Date'].dt.year
    df['Month'] = df['Date'].dt.month
    df['Day'] = df['Date'].dt.day

    df.drop('Date', axis=1, inplace=True)
    
    return df

@st.cache_resource
def load_model(model_path: str) -> KMeans:
    return joblib.load(model_path)

def process_data(input_X: pd.DataFrame) -> np.array:
    """ 
    Processes user input data into usable form for the KMeans model to predict.
    
    Args:
        input_X (pd.DataFrame): Input data in dataframe format with one instance.
    
    Returns:
        np.array: An numpy array for the KMeans model to predict
    """
    
    input_X = input_X.copy()
    
    # Split up the dates
    if 'Date' in input_X.columns:
        dates = pd.to_datetime(input_X['Date'])
        input_X['Date'] = dates

        input_X['Year'] = input_X['Date'].dt.year
        input_X['Month'] = input_X['Date'].dt.month
        input_X['Day'] = input_X['Date'].dt.day

        input_X.drop('Date', axis=1, inplace=True)
    
    input_X = pd.get_dummies(input_X, prefix=['Source'], dtype=int)
    
    for col in ['Source_GCAG', 'Source_GISTEMP']:
        if col not in input_X.columns:
            input_X[col] = 0

    # Reorder columns to ensure correct order
    input_X = input_X[['Mean', 'Year', 'Month', 'Day', 'Source_GCAG', 'Source_GISTEMP']]
   
    arr_X = input_X.to_numpy() 
    
    return arr_X
    
def plot_clusters(model: KMeans, X: np.array, input_X: np.array) -> None:
    """
    Plots the predicted class to the clusters.
    
    Args:
        model (KMeans): A KMeans model trained on X input
        X (np.array): The numpy array version of the dataset
        input_X (np.array): The numpy array of the input 
        
    Returns:
        None
    """
    centroids = model.cluster_centers_
    labels = model.labels_
    
    fig = plt.figure(figsize=(10,6))
    ax = fig.subplots()

    for cluster in range(3):
        cluster_points = X[labels == cluster]
        plt.scatter(cluster_points[:, 0], cluster_points[:, 1], label=f'Cluster {cluster}')

    # Plot centroids
    ax.scatter(centroids[:, 0], centroids[:, 1], s=200, c='black', marker='X', label='Centroids')
    
    predictions = model.predict(input_X)
    
    st.write(f"Predicted Cluster: {predictions[0]}")

    # Highlight the predicted cluster and point
    ax.scatter(input_X[:, 0], input_X[:, 1], s=300, c='red', marker='P', label=f'Predicted Cluster: {predictions[0]}')

    ax.set_title('K-Means Clustering with Predicted Point')
    ax.legend()
    st.pyplot(fig)

def main():
    st.title("Global Temperature Time Series")
    df = load_dataset("./monthly_csv.csv")
    
    tab1, tab2 = st.tabs(["KMeans Prediction", "About the Dataset"])
    
    with tab1: 
        st.header("Input Data")
        
        source = st.selectbox("Choose your source platform.", tuple(df['Source'].unique()))
        mean_temp = st.slider("Choose your avg. temp", df['Mean'].min(), df['Mean'].max())
        date = st.date_input("Choose date to monitor air quality", min_value="1980-01-01", max_value=None)
        
        
        model = None
        if st.button("Predict Input!"):
            d = pd.DataFrame({"Source": [source],
                            "Mean": [mean_temp],
                            "Date": [date]
                            })
            input_X = process_data(d)
            
            model = load_model('./models/kmeans_model.pkl')
            if model is not None:
                processed_df = process_data(df)
                plot_clusters(model, processed_df, input_X)
                
    with tab2:
        st.caption("Global Temperature Time Series. Data are included from the GISS Surface Temperature (GISTEMP) analysis and the global component of Climate at a Glance (GCAG). Two datasets are provided: 1) global monthly mean and 2) annual mean temperature anomalies in degrees Celsius from 1880 to the present.")
        
        st.write("Citation:")
        st.caption("GISTEMP: NASA Goddard Institute for Space Studies (GISS) Surface Temperature Analysis, Global Land-Ocean Temperature Index.")
        st.caption("NOAA National Climatic Data Center (NCDC), global component of Climate at a Glance (GCAG).)") 

if __name__ == "__main__":
    main()