TryMLEasy: No-Code Machine Learning for CSV Datasets

TryMLEasy is a Python-powered web application that removes the coding barrier from machine learning entirely. Whether you’re a student exploring data science concepts, a domain expert who needs quick model results, or a researcher prototyping ideas, TryMLEasy gives you a fully guided, point-and-click interface to upload a CSV dataset, preprocess it, and train production-quality ML models — all without writing a single line of code.

What is TryMLEasy?

TryMLEasy is built on Streamlit, a Python framework that turns scripts into shareable web apps. The application implements a no-code approach to machine learning: instead of importing libraries, splitting data, and calling .fit() manually, you follow a linear, wizard-style workflow through clearly labeled pages — from dataset upload, to feature selection, to model training, to results. Each step in the workflow is handled by a dedicated page (powered by st_pages), and state is preserved across pages using Streamlit’s session state. This means you can move forwards and backwards through the pipeline without losing your configuration. The guided flow encourages best practices like separating feature columns from the target, applying a scaler before training, and always holding out a test split.

TryMLEasy currently supports numerical features only. Support for categorical features is planned for a future release. Make sure your CSV dataset contains numeric columns before uploading.

Key Features

TryMLEasy packs a full ML workflow into a single browser tab, covering everything from raw data exploration through model evaluation.

No-Code GUI

Upload a CSV, configure your pipeline, and train models entirely through point-and-click controls — no Python knowledge required.

10 Traditional ML Models

Choose from six classification and four regression algorithms from scikit-learn, covering the most common supervised learning tasks.

Custom Neural Network Builder

Design a multi-layer neural network using a dynamic dataframe editor. Configure neurons and activation functions per layer, then train with custom epochs and batch size.

Preprocessing & Decomposition

Apply industry-standard scalers (Standard, MinMax, Robust, Normalization) and dimensionality reduction techniques (PCA, Kernel PCA, FastICA) before training.

Correlation Heatmap

Visualize pairwise feature correlations with an interactive heatmap powered by seaborn and Plotly, helping you identify the most informative features before training.

Multiple Evaluation Metrics

Evaluate classification models on Accuracy, F1 Score, Precision, and Recall. Evaluate regression models on MSE, Explained Variance, and Max Error — plus a predicted-vs-true scatter plot.

Supported Capabilities

TryMLEasy supports a broad and growing set of preprocessing techniques, decomposition methods, and machine learning algorithms.

Preprocessing
Decomposition
Classification Models
Regression Models
Neural Network

Scalers normalize your feature data before it reaches the model, improving convergence and preventing features with large magnitudes from dominating training.

Technique	Description
Standard Scaler	Removes the mean and scales to unit variance (z-score normalization).
MinMax Scaler	Scales each feature to a fixed range, typically [0, 1].
Robust Scaler	Scales using statistics that are robust to outliers (median and IQR).
Normalization	Scales individual samples to have unit norm (L2 by default).

Dimensionality reduction techniques transform your feature space into a lower-dimensional representation, which can reduce noise and speed up training.

Technique	Description
PCA	Principal Component Analysis — linear projection onto the axes of maximum variance.
Kernel PCA	Non-linear extension of PCA using kernel methods.
FastICA	Fast Independent Component Analysis — separates a multivariate signal into independent, non-Gaussian components.

Classification models predict a discrete class label for each input sample. All classification models are sourced from scikit-learn.

Model	Notes
Logistic Regression	Linear model for binary and multi-class classification.
Decision Tree Classifier	Tree-based model that learns decision rules from features.
Gaussian Naive Bayes	Probabilistic classifier based on Bayes’ theorem with Gaussian feature assumptions.
SVC	Support Vector Classification — finds the maximum-margin hyperplane.
K-Nearest Neighbors	Instance-based learning that classifies by majority vote among neighbors.
SGD Classifier	Stochastic Gradient Descent — efficient linear classifier for large datasets.

Regression models predict a continuous numerical value for each input sample. All regression models are sourced from scikit-learn.

Model	Notes
Linear Regression	Fits a linear relationship between features and the target.
SVR	Support Vector Regression — uses the same margin concept as SVC for continuous targets.
Ridge Regression	Linear regression with L2 regularization to reduce overfitting.
Least Angle Regression	LARS — an efficient algorithm for high-dimensional datasets.

The neural network builder lets you define a fully connected (Dense) architecture using an interactive table, then train it with TensorFlow/Keras.

Hyperparameter / Layer Type	Options
Layer type	Dense (fully connected)
Activation functions	ReLU, Sigmoid, Tanh
Neurons per layer	Any positive integer, configurable per layer
Epochs	User-defined integer
Batch size	User-defined integer
Outputs	Model summary, architecture diagram, training vs. validation loss graph, epoch-wise loss values

Tech Stack

TryMLEasy is built entirely on open-source Python libraries.

Library	Role
Streamlit	Web app framework and interactive frontend
scikit-learn	All traditional ML models, scalers, and decomposition methods
TensorFlow / Keras	Neural network definition, training, and evaluation
Matplotlib	Static plots and neural network architecture diagrams
Pandas	Dataset loading, manipulation, and display
seaborn	Correlation heatmap generation
Plotly	Interactive charts and the predicted-vs-true scatter plot

Quickstart

Launch TryMLEasy in minutes — use the live cloud app or run it locally and train your first model.

Installation

Full local setup guide: clone the repo, install dependencies, and launch the Streamlit server.

Get Started

Using TryMLEasy

Reference

Contributing

TryMLEasy: No-Code Machine Learning for CSV Datasets

What is TryMLEasy?

Key Features

No-Code GUI

10 Traditional ML Models

Custom Neural Network Builder

Preprocessing & Decomposition

Correlation Heatmap

Multiple Evaluation Metrics

Supported Capabilities

Tech Stack

Quickstart

Installation

Build docs developers (and LLMs) love

Get Started

Using TryMLEasy

Reference

Contributing

Documentation Index

​What is TryMLEasy?

​Key Features

No-Code GUI

10 Traditional ML Models

Custom Neural Network Builder

Preprocessing & Decomposition

Correlation Heatmap

Multiple Evaluation Metrics

​Supported Capabilities

​Tech Stack

Quickstart

Installation

Build docs developers (and LLMs) love

What is TryMLEasy?

Key Features

Supported Capabilities

Tech Stack