app.py

import gradio as gr
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

import seaborn as sns

df = pd.read_csv("diabetes.csv")
df.info()

fig, axs = plt.subplots(4, 2, figsize=(15, 12))
axs = axs.flatten()
sns.distplot(df["Pregnancies"], rug=True, color="#38b000", ax=axs[0])
sns.distplot(df["Glucose"], rug=True, color="#FF9933", ax=axs[1])
sns.distplot(df["BloodPressure"], rug=True, color="#522500", ax=axs[2])
sns.distplot(df["SkinThickness"], rug=True, color="#66b3ff", ax=axs[3])
sns.distplot(df["Insulin"], rug=True, color="#FF6699", ax=axs[4])
sns.distplot(df["BMI"], color="#e76f51", rug=True, ax=axs[5])
sns.distplot(df["DiabetesPedigreeFunction"], color="#03045e", rug=True, ax=axs[6])
sns.distplot(df["Age"], rug=True, color="#333533", ax=axs[7])
plt.show()

bmi_outliers = df[df["BMI"] > 40]
bmi_outliers["BMI"].shape
df["Diabetes"] = df["Outcome"].apply(lambda x: "Diabetic" if x == 1 else "Not_Diabetic")
df["BMI"] = df["BMI"].apply(lambda x: df.BMI.mean() if x > 40 else x)

from pycaret.classification import *
print("Hello 1")
#Diab = setup(data = df,target = 'Diabetes',normalize=True, session_id=1,fix_imbalance = True)
print("2")
Diab = setup(data=df,target="Diabetes",normalize=True,session_id=1,fix_imbalance=True,numeric_features=["Pregnancies"],log_experiment=True,experiment_name="diabetes",ignore_features=["Outcome"],silent=True)

best = compare_models(n_select=15)
compare_model_results = pull()

rf = create_model("rf", fold=10)

tuned_rf = tune_model(rf)

# creating a predict function to be passed into gradio UI
def predict(
    model,
    Pregnancies,
    Glucose,
    BloodPressure,
    BMI,
    SkinThickness,
    Insulin,
    Age,
    DiabetesPedigreeFunction
):

    df = pd.DataFrame.from_dict(
        {
            "Pregnancies": [Pregnancies],
            "Glucose": [Glucose],
            "BloodPressure": [BloodPressure],
            "BMI": [BMI],
            "SkinThickness": [SkinThickness],
            "Insulin": [Insulin],
            "Age": [Age],
            "DiabetesPedigreeFunction": [DiabetesPedigreeFunction]
        }
    )

    model_index = list(compare_model_results["Model"]).index(model)
    model = best[model_index]
    pred = predict_model(model, df, raw_score=True)
    return {
        "Not Diabetic": pred["Score_Not_Diabetic"][0].astype("float64"),
        "Diabetic": pred["Score_Diabetic"][0].astype("float64")
    }


import gradio as gr

model = gr.inputs.Dropdown(list(compare_model_results["Model"]), label="Models")
Pregnancies = gr.inputs.Slider(
    minimum=1, maximum=20,step=1, default=df["Pregnancies"].mean(), label="Pregnancies"
)
Glucose = gr.inputs.Slider(
    minimum=0, maximum=200,step=1 ,default=df["Glucose"].mean(), label="Glucose"
)
BloodPressure = gr.inputs.Slider(
    minimum=0, maximum=140,step=1 ,default=df["BloodPressure"].mean(), label="BloodPressure"
)
BMI = gr.inputs.Slider(minimum=0, maximum=70, default=df["BMI"].mean(), label="BMI")

SkinThickness = gr.inputs.Slider(
    minimum=0, maximum=100,step=1,default=df["SkinThickness"].mean(), label="SkinThickness"
)
Insulin = gr.inputs.Slider(
    minimum=0, maximum=1000,step=1,default=df["Insulin"].mean(), label="Insulin"
)
Age = gr.inputs.Slider(minimum=0,maximum=100,step=1,default=df["Age"].mean(), label="Age")
DiabetesPedigreeFunction = gr.inputs.Slider(
    minimum=0,
    maximum=1,
    default=df["DiabetesPedigreeFunction"].mean(),
    label="DiabetesPedigreeFunction"
)
gr.Interface(
    predict,
    [
        model,
        Pregnancies,
        Glucose,
        BloodPressure,
        BMI,
        SkinThickness,
        Insulin,
        Age,
        DiabetesPedigreeFunction
    ],
    "label",
    live=True,
).launch()