Random Forest Prediction of Cargo Ship Fuel Consumption from Speed and Wave Conditions

Project info

• Field of studyNaval Architecture and Ocean Engineering
• LevelDoctoral dissertation (S3)
• Method typeMachine Learning
• Estimated build time± 7 working days
• ToolsTreeBagger, oobError, predict (Statistics and Machine Learning Toolbox)
• PriceUS$41 - US$250

Research goal

Predict daily fuel consumption from service speed and wave height using Random Forest (TreeBagger) and assess model performance.

Variables

Example result

• 0.038OOB error
• 0.967Test accuracy
• 0.961Macro F1
• 60Trees
• 3.605Importance Service speed
• 2.525Importance Wave height

A 60-tree random forest predicts the class of Daily fuel consumption with an OOB error of 0.038 and test accuracy of 0.967 (macro F1 0.961); Service speed and Wave height prove far more important than the two noise predictors.

Computed from illustrative synthetic data, not from a real study.

About Random Forest (TreeBagger)

TreeBagger builds a random forest from many bootstrapped decision trees and gives out-of-bag error and variable importance. It suits tabular classification or regression with nonlinear relations and interactions.

Steps

1. 1Prepare features and labels, split into training and test sets
2. 2Train TreeBagger with hundreds of trees and OOBPredictorImportance
3. 3Monitor the out-of-bag error against the number of trees
4. 4Predict test data and compute accuracy and confusion matrix
5. 5Read variable importance (OOBPermutedPredictorDeltaError)
6. 6Plot OOB error and feature importance

Assumptions to check

• Training data represent the target population
• No target leakage into features
• Enough trees for the OOB error to stabilise

Reading and reporting the results

Report the OOB error, test accuracy, and the most important variables. Explain that importance reflects predictive contribution, not causation.

Example MATLAB code

A short starting script with this project's variable names. The delivered version is adapted to your data and fully commented.

rng(7);
n = 500;
service_speed = randn(n, 1);
wave_height = randn(n, 1);
x3 = randn(n, 1);
X = [service_speed wave_height x3];
daily_fuel_consumption = double(service_speed.^2 + wave_height + 0.3*randn(n, 1) > 0.8);

cv = cvpartition(daily_fuel_consumption, 'HoldOut', 0.25);
Xtr = X(training(cv), :);  ytr = daily_fuel_consumption(training(cv));
Xte = X(test(cv), :);      yte = daily_fuel_consumption(test(cv));

b = TreeBagger(200, Xtr, ytr, 'Method', 'classification', ...
    'OOBPrediction', 'on', 'OOBPredictorImportance', 'on', 'MinLeafSize', 3);
pred = str2double(predict(b, Xte));
fprintf('OOB error = %.3f\n', oobError(b, 'Mode', 'ensemble'));
fprintf('Test accuracy = %.3f\n', mean(pred == yte));
disp(confusionmat(yte, pred));
disp(b.OOBPermutedPredictorDeltaError);

figure;
subplot(1,2,1); plot(oobError(b)); xlabel('Number of trees'); ylabel('OOB error');
subplot(1,2,2); bar(b.OOBPermutedPredictorDeltaError);
set(gca, 'XTickLabel', {'x1', 'x2', 'x3'}); ylabel('Importance');

Packages and prices

Choose how much help you need. Each higher package includes everything in the lower ones. Prices exclude 11% VAT.

• Payment: 50% down payment before work starts, balance after the deliverables are handed over
• Tax: Prices exclude 11% VAT
• Data and academic integrity: Results are computed from your own research data; you write and stand behind your manuscript, we help with the analysis and your understanding of it
• Revisions: Up to 2 revisions are included in the price (Package 3: 3 rounds)

Frequently asked questions