Joe Samuel← All work

Duration Estimation for Future Canadian Armed Forces Operations

Collated and examined historical mission data using Python, R and derived insights — patterns and correlations — that helped improve Canadian Armed Forces mission planning.

The Problem

The Canadian Armed Forces required a data-driven methodology to better predict the duration of future operations based on historical mission data. Manual analysis of heterogeneous mission datasets was insufficient for systematic forecasting.

The Solution

A research report was generated for the Department of National Defence to help improve future mission planning based on data-driven insights. The project was conducted in partnership with Defence Research and Development Canada from January to April 2020.

Why It's Hard

Historical mission data was preprocessed by collating, imputing, transforming, and eliminating data from the CAF datastore. Feature selection was performed through variable importance random forest technique. A random forest classifier was developed to predict duration classes for a future mission, optimizing resources based on historical trends for similar missions grouped by region and context.

The Process

  1. Discovery

    Data Acquisition & Cleaning

    Collated, imputed, transformed, and cleaned heterogeneous historical mission data from the CAF datastore for analysis.

    Data inventoryCleaning pipeline
  2. Architecture

    Feature Selection & Model Development

    Applied variable importance random forest technique for feature selection, then developed a random forest classifier to predict mission duration classes.

    Feature importance reportModel spec
  3. Ship

    Research Report Delivery

    Generated a research report for the Department of National Defence to improve future mission planning using data-driven insights.

    Research reportDND briefing

Architecture

CAF Datastore

Historical Mission Data

Data Pipeline

SQL / R / MATLAB

Random Forest Classifier

Weka / R

Tableau Dashboard

Visualization & Reporting

CAF Datastore

Historical Mission Data

Data Pipeline

SQL / R / MATLAB

Random Forest Classifier

Weka / R

Tableau Dashboard

Visualization & Reporting

How It Works

1

Data Preparation

Historical CAF mission records are collected, cleaned, and transformed into a structured dataset suitable for ML analysis.

2

Classification

A random forest classifier groups missions by region and context, then predicts duration classes based on learned patterns.

3

Planning Insights

Results are visualized and delivered as a report, helping planners optimize resource allocation for future operations.

Techniques

  • ✓Random Forest
  • ✓Machine Learning
  • ✓Data Cleaning

Technologies

  • ✓SQL
  • ✓R
  • ✓MATLAB
  • ✓Tableau
  • ✓Weka

Previous

Merak Asset Threat Analysis Tool

Next

System-Level Security for IoT-Enabled e-Health Systems

Let's connect

EmailLinkedInGitHubGoogle ScholarResume ↓