DataSights: exploratory data analysis and machine learning, no code required

01Import

Drop a file, get a profiled dataset.

No connectors to configure, no schema to declare. A flat file in, a documented dataset out.

  • CSV, TXT, XLSX and XLS up to 100 MB, with separator and header options.
  • Validated in depth before ingestion: extension whitelist, file signature checks, size caps.
  • Instant profile: rows, columns, duplicates, memory, per-column types and statistics, with Preview and Summary Table.

Drag and drop your file here

Accepted: .csv.txt.xlsx.xls

dataset.csvReady
1,284Rows
18Columns
0Duplicates
132KB
02Prepare

Seven steps to analysis-ready.

A guided console walks you through every cleaning decision, and shows the effect before you commit.

  • Target and features validation, duplicates, types and encoding, missing values, outliers, normalization, imbalance.
  • Real methods: kNN imputation, IQR and z-score detection, winsorization, robust scaling, one-hot and frequency encoding, SMOTE, ADASYN and Tomek resampling.
  • Before and after visuals at each step, and the processed dataset downloadable at any point.
Preprocessing console
areaQuantitative812 unique
roomsQuantitative9 unique
ageQuantitative64 unique
Duplicates Detection
Rows before200
Duplicates found20
Rows after180
03Explore

Every distribution, every correlation.

Univariate, bivariate and correlation analysis with real chart controls, expert warnings and honest statistics.

  • Histograms, boxplots, scatter, heatmaps with regression line, confidence bands and bin controls.
  • Key metrics with plain-language interpretation, a full distribution table, and methodological warnings that flag skew, small samples and unreliable regions.
  • Your exploration journey is saved: the charts you generate land in the report and the notebook.
Histogram + Densityprice · 1,284 obs
Central Tendency
6.285Mean
6.208Median
Mean above median: a longer upper tail pulls the mean up.
04Model

Models you can actually defend.

Preconditions checked, evaluation chosen deliberately, metrics reported with their uncertainty.

  • 12 algorithms across classification and regression, hyperparameters included, optional Grid Search tuning.
  • Honest evaluation, five ways: holdout, random subsampling, K-Fold, stratified K-Fold, leave-one-out, always on held-out data.
  • Metrics with mean and std, confusion matrix, ROC curve, feature importances, and a saved .pkl.
TrainingK-Fold (5)
Random ForestDone
Linear RegressionTraining
Model results
0.624R² ± 0.041
0.431RMSE ± 0.028
0.312MAE ± 0.019
05Predict

Raw values in, predictions out.

The fitted preprocessing pipeline is replayed at inference: you type raw values, DataSights handles the rest.

  • Single and batch modes: a typed form pre-filled with sensible defaults, or a CSV and Excel upload.
  • Column mapping when your file names differ from the training features.
  • Compare several models at once, probabilities included, and export predictions as CSV.
Predict · Singleraw inputs
84
3
12
Random Forest
231,900
predicted price
06Report

A document that writes itself.

Every step you took, every chart you generated, assembled into a report and a runnable notebook.

  • Auto-built sections from your actual pipeline: profile, preprocessing summary with before and after, Explore charts, model results. Edited inline, block by block.
  • Three PDF templates with accent colors, vector charts, in English, French or Spanish. Technical ships a reproducibility annex.
  • Notebook export (.ipynb), trilingual, that reproduces your whole journey, charts included.
Report settings
Template
ExecutiveTechnicalStudio
Accent

PDF up to date · generated at 14:02

Your data has a story.Tell it end to end.

From a raw file to a report you can defend, in one sitting.

Start analyzing