iOS Machine Learning Deployment with Core ML and Vapor

Load the Carvana csv with pandas in a Jupyter notebook, inspect the data frame, check for nulls, and filter years above 2025 to identify corrupted rows and plan fixes.

Fix the year column using Python and pandas by converting to string, selecting four characters, casting to integer, and reassigning to the data frame; verify with min, max, and head.

Standardize the miles column using sklearn's standard scalar with fit_transform to align scales, compute mean and standard deviation, and prepare standardized values for model training.

Learn how to convert the name column from strings to numbers using label encoding and why it pairs well with random forest, as an alternative to one hot encoding.

Learn linear regression, a supervised method that predicts a continuous value with a linear model y = b0 + b1 x, learned by ordinary least squares, evaluate MAE, RMSE, R^2.

Explore how the random forest ensemble builds multiple decision trees, uses bagging and feature randomness to reduce overfitting and improve accuracy for both classification and regression tasks.

Explore one hot encoding to convert categories into binary columns that models can process. Compare with label encoding to avoid ordinal biases and learn when to use get_dummies in practice.

Encode text categories into numbers to prepare data for algorithms, including tree-based models like random forests. Prefer one-hot encoding when needed to avoid introducing artificial order.

Train a random forest regressor on the dataset by encoding features, performing an 80/20 train-test split, and evaluating with mean absolute error and R2 score for deployment readiness.

Convert the sklearn model to a core ml model with coremltools and save it as carvana.mlmodel for iOS deployment; host on vapor and expose a JSON API for updates.

Create a car name mapping with python dictionary comprehension and zip to pair unique car names with label-encoded values, then save as a json file for the iOS app.

Populate car names from a json file into a picker, map names to IDs, and build a form with year and miles for price prediction.

Integrate Carvana Core ML model into the iOS app by dragging the model and generating Swift model class. Wire year, miles, and encoded name to predict price.

Discover how standardizing miles with a mean and standard deviation improves car price predictions. See how Vapor enables server hosting of the model for API updates without redeploying the app.

Install Vapor on macOS, create a hello world Vapor project, run the local server with root and /hello routes, and prepare for Core ML model integration.

Learn how to integrate a Core ML model into a Vapor app, compile it to a Carvana file, organize resources, and troubleshoot bundle loading.

Develop a thread safe car price predictor as a Vapor actor that loads a Carvana Core ML model, wraps it, and exposes a post route with request and response DTOs.

Integrate a Vapor-hosted model with a SwiftUI iOS app by building an HTTP client to post price prediction requests with miles, year, and name encoded value, returning a JSON price.

Deploy your Vapor app locally and publicly expose its JSON API with Ngrok, by installing Ngrok, running the Vapor server on 8080, and testing via Postman.