gradio/demo/fraud_detector.py

# Demo: (Timeseries, CheckboxGroup, Slider) -> (Timeseries, Dataframe, Label)

import gradio as gr
import pandas as pd
import random


def fraud_detector(card_activity, categories, sensitivity):
    activity_range = random.randint(0, 100)
    drop_columns = [column for column in ["retail", "food", "other"] if column not in categories]
    if len(drop_columns):
        card_activity.drop(columns=drop_columns, inplace=True)
    return card_activity, card_activity, {"fraud": activity_range / 100., "not fraud": 1 - activity_range / 100.}


iface = gr.Interface(fraud_detector,
                     [
                         gr.inputs.Timeseries(
                             x="time",
                             y=["retail", "food", "other"]
                         ),
                         gr.inputs.CheckboxGroup(["retail", "food", "other"], default=[
                             "retail", "food", "other"]),
                         gr.inputs.Slider(1, 3)
                     ],
                     [
                         "dataframe",
                         gr.outputs.Timeseries(
                             x="time",
                             y=["retail", "food", "other"]
                         ),
                         gr.outputs.Label(label="Fraud Level"),
                     ]
                     )
if __name__ == "__main__":
    iface.launch()