Making Machine Learning Accessible with Streamlit

The Problem with Machine Learning

Machine learning is powerful, but it has a massive accessibility problem. To run even a simple classification model, you typically need to: install Python, set up a virtual environment, write data loading code, preprocess the data, train the model, and write evaluation code.

That's a huge barrier for business analysts, students, and domain experts who understand their data but don't know Python.

My ML App removes that barrier entirely.

What Streamlit Makes Possible

Streamlit is remarkable. You write a regular Python script, and Streamlit turns it into an interactive web app automatically. No HTML, no CSS, no JavaScript:

import streamlit as st
import pandas as pd
from sklearn.ensemble import RandomForestClassifier

st.title("Train Your ML Model")

uploaded_file = st.file_uploader("Upload your CSV", type="csv")
if uploaded_file:
    df = pd.read_csv(uploaded_file)
    st.dataframe(df.head())
    
    target = st.selectbox("Select target column", df.columns)
    algorithm = st.selectbox("Select algorithm", 
                             ["Random Forest", "Logistic Regression", "SVM"])
    
    if st.button("Train Model"):
        X = df.drop(target, axis=1)
        y = df[target]
        # ... training logic

The Full Feature Set

My ML App supports:

• Upload any CSV dataset: Automatic column type detection
• Multiple algorithms: Random Forest, Logistic Regression, Decision Tree, SVM, KNN
• Automatic preprocessing: Handles missing values, encodes categorical variables
• Visual evaluation: Confusion matrix, ROC curve, feature importance charts
• Model comparison: Train multiple models and compare their metrics side-by-side
• Predictions: Upload new data and get predictions from your trained model

The Preprocessing Pipeline

Getting preprocessing right is crucial. I built a pipeline that handles the messy reality of real-world datasets:

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler, LabelEncoder
from sklearn.impute import SimpleImputer

def build_pipeline(algorithm):
    return Pipeline([
        ('imputer', SimpleImputer(strategy='median')),
        ('scaler', StandardScaler()),
        ('model', get_model(algorithm))
    ])

Using scikit-learn's Pipeline means preprocessing and the model are treated as a single unit — you can't accidentally apply different preprocessing to your training and test data.

Deployment on Streamlit Cloud

Streamlit Cloud is the easiest way to deploy a Streamlit app. Push your code to GitHub, connect your repo, and your app is live in under 2 minutes. Free tier is more than enough for most projects.

Try the live app and explore the source code.