Machine Learning Forecast Lab

Predicting
Food Prices
Ahead.

Move from historical analysis to forward-looking estimates. This page presents model evaluation, forecast behavior, uncertainty, and crop-level signals in one interactive forecasting workspace.

₹78.74
MAE
₹134.46
RMSE
1.99%
MAPE
0.9941
R²

Forecast engine

Extra Trees

Evaluated

RMSE

₹134.46

MAE

₹78.74

MAPE

1.99%

R²

0.9941

Test

243

Features

58

₹3,888₹1,944₹0
ActualForecast

Historical test-set predictions from the exported Extra Trees evaluation output.

01 / Forecast trajectory

Actuals meet the forecast.

Actual and predicted prices from the 243-observation chronological holdout used to evaluate the Extra Trees model. Select a crop to inspect its month-by-month test predictions.

27-observation test view

Rice — actual vs predicted

Extra Trees
₹3,888₹2,916₹1,944₹972₹0Sep 2023Oct 2024Nov 2025
Sep 2023Nov 2025
Observed price
Latest test point
₹3,500Nov 2025
Model prediction
Extra Trees
₹3,514Nov 2025

02 / Model evaluation

Measure the model, not just the prediction.

Model quality is evaluated using regression metrics rather than a generic “accuracy” percentage. Lower MAE/RMSE and lower MAPE indicate smaller errors; higher R² indicates stronger explained variance.

MAE

₹78.74

Average absolute error

RMSE

₹134.46

Penalizes larger errors

MAPE

1.99%

Relative error

R²

0.9941

Explained variance

Benchmark table

Compare forecasting approaches

ModelRMSEMAEMAPESMAPER²Status
Extra Trees₹134.46₹78.741.99%1.97%0.9941Selected
Hist Gradient Boosting₹211.62₹99.871.96%1.92%0.9855Comparison
Random Forest₹346.30₹231.807.22%7.08%0.9611Comparison
Naive Last Month₹458.69₹271.2210.07%—0.9317Baseline

03 / Crop performance explorer

Compare the signals. See where errors differ.

These metrics are calculated from the same chronological test predictions used in the overall evaluation.

01 / commodity

Maize

27 test

MAE

₹5

RMSE

₹6

MAPE

0.24%

R²

0.9985

02 / commodity

Wheat

27 test

MAE

₹7

RMSE

₹10

MAPE

0.29%

R²

0.9956

03 / commodity

Rice

27 test

MAE

₹18

RMSE

₹21

MAPE

0.53%

R²

0.9842

04 / commodity

Potato

13 test

MAE

₹32

RMSE

₹38

MAPE

3.17%

R²

0.9777

05 / commodity

Jowar/Sorghum

27 test

MAE

₹78

RMSE

₹98

MAPE

2.77%

R²

0.9609

06 / commodity

Onion

27 test

MAE

₹95

RMSE

₹148

MAPE

3.25%

R²

0.9796

07 / commodity

Banana

27 test

MAE

₹99

RMSE

₹129

MAPE

2.67%

R²

0.9447

08 / commodity

Cotton

27 test

MAE

₹106

RMSE

₹145

MAPE

1.51%

R²

0.5611

09 / commodity

Groundnut

27 test

MAE

₹278

RMSE

₹309

MAPE

4.84%

R²

0.3812

03.5 / Forecast analytics lab

Interrogate the forecast. Not just view it.

Additional interactive views calculated directly from the exported chronological test predictions and verified model evaluation metrics.

Error trajectory

Rice · absolute percentage error

27 observations
0%0%1%1%2%
Sep 2023chronological test windowNov 2025

Selected crop signal

Rice

RMSE on chronological holdout

₹21

Error distribution

Rice percentage-error buckets

max 1.55%
<1%23
1–2%4
2–4%0
4–6%0
6–8%0
8%+0

Model benchmark

Compare evaluation metrics

Final evaluation: 243 chronological test observations across 9 complete forecasting crops. Extra Trees was selected by lowest test RMSE.

Extra Trees₹134.46
Hist Gradient Boosting₹211.62
Random Forest₹346.30
Naive Last Month₹458.69

Prediction telemetry

Observed vs predicted snapshot

Extra Trees
Jun 2025

₹3,416

err 0.46%

Jul 2025

₹3,436

err 0.46%

Aug 2025

₹3,421

err 0.32%

Sep 2025

₹3,447

err 0.03%

Oct 2025

₹3,425

err 0.74%

Nov 2025

₹3,514

err 0.40%

observedpredictedlatest ₹3,514
\n\n

04 / Residual behavior

Where does the model miss?

Residuals help reveal whether errors are centered around zero or whether the model repeatedly under- or over-estimates particular periods.

Test observations

243

Chronological holdout observations used for evaluation.

Predictor columns

58

Model features used by the forecasting pipeline.

05 / Forecast dashboard

One place for the forecast story.

This dashboard frame is prepared for the final Tableau and Power BI integration. The page already exposes the model metrics, crop filters, sorting, forecast curve, and error diagnostics needed for the interactive layer.

Forecast monitor

Observed vs predicted

R² 0.9941
₹3,888₹1,944₹0

Model metrics

MAE

₹78.74

RMSE

₹134.46

MAPE

1.99%

Pipeline

✓Processed data
✓Feature engineering
✓Chronological split
✓Model evaluation
✓Forecast evaluation

Next layer

Tableau

Next layer

Power BI

Data source

Processed CSV

Tableau Public · Food Price Forecasting

Interactive Tableau Dashboard

Open Tableau ↗Live Report
Responsive Tableau view100%

Power BI · Food Price Forecasting

Interactive Power BI Forecasting Report

Forecasting workflow

From cleaned records to future estimates.

The forecasting layer sits after ingestion, cleaning, EDA, and feature engineering. The next step is to connect the final exported predictions so every visual on this page is driven directly by the project artifacts.

01

Prepare

Temporal features and lag variables

02

Train

Chronological train-test split

03

Evaluate

MAE · RMSE · MAPE · R²

04

Forecast

Generate future price estimates