GitHUB :https://github.com/RamyaNandhan/machineLearning/tree/main/decision-tree-ml-project

This is an example of classical Machine Learning pipeline built with:
Visualization / EDA: seaborn, matplotlib,
Modeling: scikit-learn,
Model: DecisionTreeClassifier,
Tuning: GridSearchCV,
Evaluation: sklearn.metrics,
Data split: train_test_split,
Storage source: Google Drive (Colab)

Machine Learning_Architecture :

Microservice level Machine Learning Architecture

The microservice_ml_architecture diagram includes these services:

Client Applications

API Gateway

Prediction (Inference) Service

Data Ingestion Service

Feature Engineering Service / Feature Store

Model Training Service

Hyperparameter Tuning Service

Model Evaluation Service

Model Registry

Data Storage

Why Microservice ML Architecture?

Advantages

Independent scaling of services

Easier model updates

Fault isolation

Continuous deployment of models

Works well with MLOps pipelines

Data Flow

User sends request to API Gateway

Request routed to Prediction Service

Prediction service retrieves features

Latest model fetched from Model Registry

Prediction returned to client

Data logged for future training pipeline

The Problem on this notebook, is basically a comparison of three Decision Tree models:

Default Decision Tree

Pre-Pruned Decision Tree

Post-Pruned Decision Tree

we get to evaluate them using Accuracy, Recall, Precision, and F1 score.

This article explains a classification model using a Decision Tree to predict a target category based on input features.
Typical real-world problems that use this exact approach include:

Customer churn prediction

Loan approval / credit risk

Fraud detection

Medical diagnosis classification

Customer purchase prediction

demonstrates the complete ML lifecycle:

Data loading, Data exploration,Model training,Hyperparameter tuning,Model evaluation,Selecting the best model

Overall, this article shows steps to build a machine learning classification pipeline using a Decision Tree model. The workflow included data preprocessing, exploratory analysis, model training, hyperparameter tuning using GridSearchCV, and evaluation using accuracy, precision, recall, and F1 metrics.
compared default, pre-pruned, and post-pruned trees and selected the post-pruned model because it achieved the best balance between accuracy and generalization.