My First ML Web App: Diabetes Prediction with Streamlit (Mistakes, Fixes & Lessons Learned)

Introduction

When I started my college summer internship, my goal was simple — build something real with machine learning.
I decided to create a Diabetes Prediction Web App using Python and Streamlit.

Spoiler: I had no idea that I’d spend hours fighting with GitHub errors like "fetch first" and confusing bugs like ModuleNotFoundError: joblib.
But that’s the beauty of projects — they teach you way more than just coding.

In this blog, I’ll Walk you through how I built this app from scratch, the mistakes I made, and what I learned.

Step 1: Choosing the Problem

I picked the PIMA Indians Diabetes Dataset because:

• It’s simple yet meaningful.

• It predicts whether a person is diabetic based on features like glucose, BMI, and age.

I downloaded diabetes.csv and explored it using Pandas:

pythonCopy codeimport pandas as pd

data = pd.read_csv("diabetes.csv")
print(data.head())

---

Step 2: Building the ML Model

For this project, I chose KNN (K-Nearest Neighbors). Why?
Because KNN is easy to understand — it predicts based on the closest neighbors in data.

The training code was as simple as:

pythonCopy codefrom sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
import joblib

X = data.drop('Outcome', axis=1)
y = data['Outcome']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

model = KNeighborsClassifier(n_neighbors=5)
model.fit(X_train, y_train)

# Save model
joblib.dump(model, "Diabetes_KNN.pkl")

---

Step 3: Adding a UI with Streamlit

I didn’t want this to remain a boring .py script.
So, I used Streamlit to make a simple web interface.

Here’s how my app.py looked:

pythonCopy codeimport streamlit as st
import joblib
import numpy as np

model = joblib.load("Diabetes_KNN.pkl")

st.title("Diabetes Prediction App")
glucose = st.number_input("Glucose Level")
bmi = st.number_input("BMI")
age = st.number_input("Age")

if st.button("Predict"):
    prediction = model.predict([[glucose, bmi, age, ...]])
    st.success("Diabetic" if prediction[0] == 1 else "Not Diabetic")

---

Step 4: My Mistakes (and Fixes)

Here’s where the real learning started:

Mistake 1: Wrong requirements file

• I created requirement.txt instead of requirements.txt.

• Fix: Renamed the file and added joblib:

    nginxCopy codestreamlit
    scikit-learn
    pandas
    joblib
  

Mistake 2: GitHub Push Errors

• I kept seeing:

    vbnetCopy codeerror: failed to push some refs to 'github.com:AstutiJ/Diabetes-detect.git'
  

• Fix: Learned to git pull --rebase before pushing:

    bashCopy codegit pull origin main --rebase
    git push origin main
  

Mistake 3: ModuleNotFoundError

• On Streamlit Cloud, I got:

    makefileCopy codeModuleNotFoundError: joblib
  

• Fix: Adding joblib to requirements solved it.

---

Step 5: Deployment

I connected my GitHub repo to Streamlit Cloud, and within minutes, my app was live.
This was the best feeling ever — sharing my app with others!

---

Lessons Learned

• Building projects isn’t just about code — it’s about debugging, experimenting, and learning.

• I now understand Git basics, ML model saving, and web app deployment.

---

Conclusion

This project was part of my college summer internship program, and it helped me connect machine learning with real-world applications.
Next, I’m planning to work on a Breast Cancer Detection App (stay tuned — I’ll share that journey too!).

---

GitHub Repository: Diabetes Detection App