app.py

import gradio as gr
import pandas as pd
import numpy as np
from scipy.signal import find_peaks
from statsmodels.tsa import stattools
import plotly.graph_objects as go
import matplotlib.pyplot as plt


MIN_ALPHA = 0.01
MAX_ALPHA = 0.2
STEP = 0.01

target_names = {"SeoulBikeData": "Rented Bike Count",
                "daily-total-female-births": "Births",
                "southern_oscillation_evenly_spaced": "oscillation",
                "monthly_co2": "CO2",
                "monthly-sunspots": "Sunspots",
                "daily_bike_shares": "cnt",
                "AirPassengers": "#Passengers",
                "Alcohol_Sales": "S4248SM144NCEN",
                "yahoo_stock": "Adj Close",
                "Miles_Traveled": "TRFVOLUSM227NFWA",
                "DailyDelhiClimateTrain": "meantemp",
                "peyton_manning": "y"}



def plot_significant_lags(conf_level: float, dataset: str):
    df = pd.read_csv(dataset + ".csv")
    y = df.pop(target_names[dataset])
    acf_values, ci_intervals = stattools.acf(y, nlags=min(len(y) - 1, 400), fft=True, alpha=conf_level)
    peaks, _ = find_peaks(acf_values)
    index = np.arange(len(acf_values))
    significant = np.logical_or(ci_intervals[:, 0] > 0, ci_intervals[:, 1] < 0)
    first_significant_10 = index[:10][significant[:10]]
    significant_lags = sorted(set(index[significant]).intersection(peaks).union(first_significant_10))
    not_significant_lags = sorted(set(index).difference(significant_lags))
    assert not set(significant_lags).intersection(not_significant_lags)
    
    fig1 = go.Figure()
    fig1.add_trace(go.Scatter(x=significant_lags, 
                              y=acf_values[significant_lags],
                          mode = 'markers',
                          marker_color ='red',
                          marker_size  = 10))
    fig1.add_trace(go.Scatter(x=not_significant_lags, 
                              y=acf_values[not_significant_lags],
                          mode = 'markers',
                          marker_color ='black',
                          opacity=0.2,
                          marker_size  = 10))
    fig1.add_trace(go.Scatter(x=np.arange(len(acf_values)), y=ci_intervals[:, 0] - acf_values,   
                              line_color='blue', opacity=0.1))
    fig1.add_trace(go.Scatter(x=np.arange(len(acf_values)), y=ci_intervals[:, 1] - acf_values,
                              line_color='blue', opacity=0.1, fill='tonexty'))
    fig1.update_yaxes(range=[min(-0.2, min(acf_values), min(ci_intervals[:, 0])) - 0.1, 1.1])
    fig1.update_layout(showlegend=False,
                      title_text=f"{dataset}<br>Statistically significant lags for a {int((1 - conf_level) * 100)}% confidence interval:<br> {significant_lags}")
    return fig1


demo = gr.Blocks()

with demo:
    with gr.Column():
        conf_level_slider = gr.Slider(minimum=MIN_ALPHA, maximum=MAX_ALPHA, step=STEP)
        dataset = gr.Dropdown(choices=["SeoulBikeData", "montly_co2", "daily_bike_shares", "peyton_manning"],
        value="peyton_manning")
        inputs = [conf_level_slider, dataset]
    with gr.Column():
        plot = gr.Plot()
        button = gr.Button("Visualize")
        button.click(plot_significant_lags, inputs=inputs, outputs=plot)

demo.launch()