Decision Intelligence

Model strategy before model training.

Upload a CSV, select the target if needed, and Quantsynth will organize the dataset, run EDA, and recommend concrete models, preprocessing, and initial hyperparameters with clear reasoning.

Works for both time series and tabular datasets.
Keeps the analysis legible, grouped, and downloadable.
Prioritizes explainability over opaque output.

Generate a recommendation

The only required file is the CSV. Quantsynth detects whether the dataset is time series or tabular, but you can override the target column whenever the file is ambiguous.

No account required. Anonymous access includes 5 successful tool runs per rolling week across Quantsynth tools.

No file selected

Try a real grouped forecasting dataset or a real churn-classification dataset before uploading your own file.

What the report returns

An executive decision view with dataset type, target, first model, confidence, health, and critical warnings.
A ranked model path with baseline, challengers, validation strategy, and calibrated reasoning.
Target selection, leakage risk, preprocessing, feature engineering, EDA, plots, and detailed diagnostics in separate sections.
Structured downloads that preserve the same analysis, confidence fields, and report organization.

Coverage Time series and tabular

Output style Readable, grouped, explainable

This is advisory by design. The goal is to help the user understand the dataset and choose the next experiment with confidence.

Decision Intelligence Report

Decision Summary

Decision first. Read this section to understand the dataset and first model path in a few seconds.

Dataset Health & Risk

Assess whether this dataset is reliable enough for a first modeling pass.

Dataset Health

Target Variable

Define exactly what is being predicted and how confidently the engine selected it.

Recommended Model Path

Start with the baseline, benchmark the challengers, and validate using the right split strategy.

Model candidates

Validation strategy

Modeling workflow

Initial hyperparameters

Feature Engineering

Separate retained signal from features to create later, deprioritize, or handle cautiously.

Keep

Create

Deprioritize

Exclude / Risky

Column-specific ideas

Preprocessing & Imputation

A compact summary first, then the per-column preparation details.

EDA & Signal Analysis

Evidence comes after the recommendation path, not before it.

Signal overview

Top predictive features

Mutual information

Interaction potential

Plots

Visual evidence grouped into target, feature, and structural views.

Feature Profiles

Reference material for per-feature stats and preparation choices.

Detailed Diagnostics

Raw diagnostic depth for integrity, sparsity, cleaning, and schema consistency.