app.py

import streamlit as st
import pandas as pd
import numpy as np
from prophet import Prophet
from prophet.diagnostics import performance_metrics
from prophet.diagnostics import cross_validation
from prophet.plot import plot_cross_validation_metric
import base64

st.title('📈 Automated Time Series Forecasting')

"""
This data app uses Facebook's open-source Prophet library to automatically generate future forecast values from an imported dataset.
You'll be able to import your data from a CSV file, visualize trends and features, analyze forecast performance, and finally download the created forecast 😵 

**In beta mode**

Created by Zach Renwick: https://twitter.com/zachrenwick

Code available here: https://github.com/zachrenwick/streamlit_forecasting_app
"""

"""
### Step 1: Import Data
"""
df = st.file_uploader('Import the time series csv file here. Columns must be labeled ds and y. The input to Prophet is always a dataframe with two columns: ds and y. The ds (datestamp) column should be of a format expected by Pandas, ideally YYYY-MM-DD for a date or YYYY-MM-DD HH:MM:SS for a timestamp. The y column must be numeric, and represents the measurement we wish to forecast.', type='csv')

st.info(
            f"""
                👆 Upload a .csv file first. Sample to try: [peyton_manning_wiki_ts.csv](https://raw.githubusercontent.com/zachrenwick/streamlit_forecasting_app/master/example_data/example_wp_log_peyton_manning.csv)
                """
        )

if df is not None:
    data = pd.read_csv(df)
    data['ds'] = pd.to_datetime(data['ds'],errors='coerce') 
    
    st.write(data)
    
    max_date = data['ds'].max()
    #st.write(max_date)

"""
### Step 2: Select Forecast Horizon

Keep in mind that forecasts become less accurate with larger forecast horizons.
"""

periods_input = st.number_input('How many periods would you like to forecast into the future?',
min_value = 1, max_value = 365)

if df is not None:
    m = Prophet()
    m.fit(data)

"""
### Step 3: Visualize Forecast Data

The below visual shows future predicted values. "yhat" is the predicted value, and the upper and lower limits are (by default) 80% confidence intervals.
"""
if df is not None:
    future = m.make_future_dataframe(periods=periods_input)
    
    forecast = m.predict(future)
    fcst = forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']]

    fcst_filtered =  fcst[fcst['ds'] > max_date]    
    st.write(fcst_filtered)
    
    """
    The next visual shows the actual (black dots) and predicted (blue line) values over time.
    """
    fig1 = m.plot(forecast)
    st.write(fig1)

    """
    The next few visuals show a high level trend of predicted values, day of week trends, and yearly trends (if dataset covers multiple years). The blue shaded area represents upper and lower confidence intervals.
    """
    fig2 = m.plot_components(forecast)
    st.write(fig2)


"""
### Step 4: Download the Forecast Data

The below link allows you to download the newly created forecast to your computer for further analysis and use.
"""
if df is not None:
    csv_exp = fcst_filtered.to_csv(index=False)
    # When no file name is given, pandas returns the CSV as a string, nice.
    b64 = base64.b64encode(csv_exp.encode()).decode()  # some strings <-> bytes conversions necessary here
    href = f'<a href="data:file/csv;base64,{b64}">Download CSV File</a> (right-click and save as ** &lt;forecast_name&gt;.csv**)'
    st.markdown(href, unsafe_allow_html=True)