01 / LOOK⌕
Meet your data
See columns, data types, missing values, duplicates, summaries, and correlations in one place.
daytotal_billtip
Sun16.991.01
Sat10.341.66
Thur21.013.50
OPEN SOURCE · BUILT FOR LEARNING
Make data
make sense.
From a CSV you don’t understand to a machine learning experiment you can explain. DataForge helps you explore, learn, and build—one clear step at a time.
your_data.csv244 rows · 7 columns
DATA
→ INSIGHT
ExplorePatterns, made visible
Try a modelLearn by experimenting
A LITTLE TOOL FOR BIG “AHA” MOMENTS
SCROLL TO EXPLORE01 — THE IDEA
INSPECT YOUR DATA ✳LEARN MACHINE LEARNING ✳BUILD REAL NOTEBOOKS ✳KEEP YOUR DATA LOCAL ✳INSPECT YOUR DATA ✳LEARN MACHINE LEARNING ✳BUILD REAL NOTEBOOKS ✳KEEP YOUR DATA LOCAL ✳
01THE SHORT VERSION
Data science feels less intimidating when you can see each step.
DataForge is a Python toolkit with an optional local dashboard. Bring a CSV or Excel file, learn what’s inside, try a model, and take your work with you as an editable Jupyter notebook.
You choose what to explore and what to predict. DataForge checks the experiment before you download it, and explains the results in a workflow you can inspect.
Get set up in a few minutes →02 / TRY↗
Experiment safely
Preview cleaning, make a chart, and test an algorithm against sample data or your upload.
03 / TAKE IT↓
Keep learning
Download an editable notebook. Open it in Jupyter and run through every step yourself.
01 data = Dataset("tips.csv")02 data.explore()03 model.fit(X_train, y_train)▍
02THE ML PLAYGROUND
Learn by trying.
Not guessing.
Choose sample data to explore an idea, or use your uploaded CSV. Pick a model, choose what it should predict, then compare the results with charts and simple metrics.
Explore the algorithms ↓✳No magic score. See what the model got right, where it struggled, and which data it used.
DATAFORGE / PLAYGROUND READY
ILLUSTRATIVE PREVIEWCLASSIFICATION
What patterns can a model find?
DATA SOURCESample data ⌄
ALGORITHMLogistic Regression ⌄
WHAT TO PREDICTYour chosen label ⌄
1.00.50.0
EXAMPLE SCORE82%illustrative only
✓ You choose the targetEDITABLE
A SAFE PLACE TO ASK “WHAT IF?”
03WHAT’S INSIDE
Everything has
a next step.
A friendly front door for your data, backed by familiar Python tools you can keep using anywhere.
Data overview
Inspect shape, schema, missingness, duplicates, statistics, and correlations.
01Careful cleaning
Preview changes before applying them. Reset to the uploaded original whenever you need.
02Exploratory charts
Build histograms, boxplots, scatterplots, heatmaps, and category charts.
03Five algorithms
Try linear regression, logistic regression, decision tree, KNN, or K-means.
04Notebook builder
Test selections, download editable code, then run it step by step in Jupyter.
05Python library
Use DataForge directly in your own scripts and build on the same core tools.
0604THE ALGORITHM SHELF
Pick the right
kind of question.
The target is the answer you want a model to learn. Choose an algorithm that matches the kind of answer you have.
?
Quick tip
Predicting a number? Try regression. Predicting a label or group? Try classification. Finding natural groups without a target? Try K-means.
Linear Regression NUMBER
Predict a numeric value, like a bill total or house price.
Logistic Regression LABEL
Choose between labels, like yes/no or one of several categories.
Decision Tree LABEL
Make a label prediction through a sequence of learned decisions.
K-Nearest Neighbors LABEL
Predict a label from the most similar examples nearby.
K-means GROUPS
Discover groups in numeric data. It doesn’t need a target column.
05YOUR FIRST FIVE MINUTES
Bring a file.
Leave with a plan.
DataForge is a Python package. Install it once, launch the dashboard from a terminal, and open the local address it prints.
INSTALL THE DASHBOARD
python -m pip install "dataforge-sba[dashboard,visualization]"START THE DASHBOARD
dataforge-sba dashboardWindows: if the command isn’t recognized, close and reopen your terminal after installation.
Install DataForge
Use the command above in PowerShell, Terminal, or your environment’s terminal.
Open the local dashboard
Run the launch command. Visit the local address shown in the terminal.
Upload CSV or Excel
Start with a copy of a file. DataForge works with CSV and .xlsx uploads.
Explore, then experiment
Review your columns. Preview cleaning. Pick the data source, algorithm, and target.
Take it into Jupyter
Test your notebook selections, download the notebook, then open it in Jupyter.
FIRST_LOOK_AT_TIPS.IPYNB•••
MARKDOWN## What can we learn
## What can we learn
from this dataset?
PYTHON · IN [1]
import pandas as pd
df = pd.read_csv("tips.csv")
print(df.describe())total_bill 19.79
tip 2.99✓
tip 2.99✓
YOUR DATA.
YOUR CODE.
YOUR NEXT IDEA.
YOUR CODE.
YOUR NEXT IDEA.
NOTEBOOKSNO BLACK BOXES
A notebook you
can actually learn from.
The generated Jupyter notebook is editable Python—not a report locked inside the app. It lays out the selected data prep, charts, features, target, and experiment so you can read and change the code.
Move the notebook and dataset together, or update the data path at the top. Then run the cells in Jupyter to see the work happen.
What is Jupyter? ↗06LOCAL BY DESIGN
Your data stays
your data.
DataForge’s dashboard runs on your own computer. It doesn’t send your dataset to DataForge or an external service, and it doesn’t overwrite the original file.
Read the source on GitHub ↗Local dashboardOpens on your computer at 127.0.0.1.ON DEVICE
Original stays safeCleaning happens in the session; you can reset.REVERSIBLE
You control the notebookDownloaded code runs when you choose in Jupyter.YOUR CALL
The dashboard allows files up to 25 MB. Uploaded files are stored in the local .dataforge folder for your session. Remove uploads in the dashboard when done. See the README for cleanup details and safe handling tips.
YOUR TURNTHE FIRST STEP IS A CSV
Curiosity looks
good on you.
Start small. Ask one question. See what your data has to say.
MADE TO MAKE THE FIRST STEP EASIER