Click to view system information

Guided-Adaptive Smoothing Forecasting System

An Error-Driven Hybrid SES–HWES Framework for Multi-Path Range Forecasting in Uncertain Time Series

BS Computer Science Thesis | Isabela State University

Start Forecasting Simulation

Thesis Overview

Academic Research Implementation

This system represents the practical implementation of a thesis presented to the College of Computing Studies, Information and Communication Technology at Isabela State University - Echague Campus.

Research Title

Guided-Adaptive Smoothing: An Error-Driven Hybrid SES–HWES Framework for Multi-Path Range Forecasting in Uncertain Time Series

Authors

  • Jherson Aguto
  • Mark John Panganiban

Thesis Advisor

Dr. Ricardo Q. Camungao

Research Objectives

  • Develop an adaptive algorithm framework that dynamically assigns weights to SES and HWES forecasts and generates a multi-path range prediction
    • Weighted HWES Path
    • Weighted SES Path
    • Combined GAS Prediction (average of Hwes and SES predicted values)
  • Evaluate the model on both stationary and non-stationary datasets (with and without log transformation) and benchmark its performance against standalone SES and HWES
    • Mean Squared Error (MSE)
    • Mean Absolute Error (MAE)
    • Root Mean Squared Error (RMSE)
    • Mean Absolute Percentage Error (MAPE)
  • Develop a forecasting simulator capable of producing real-time multi-path predictions and displaying residual computation tables for all prediction paths to support model interpretability and validation.

Technical Architecture

ASP.NET Core WebAPI C# Backend Microsoft SQL Server Vanilla JavaScript Chart.js RESTful API
Time Series Analysis

GAS Algorithm Flowchart

GAS Algorithm Flowchart

Figure: Guided-Adaptive Smoothing (GAS) Algorithm Flowchart demonstrating the parallel execution of SES and HWES with dynamic inverse-error weighting

Algorithm Innovation

The GAS framework introduces a novel hybrid approach that:

  • Executes SES and HWES in parallel on training data (75%)
  • Applies grid-search for parameter optimization (α, β, γ)
  • Computes dynamic inverse-error weights based on MSE performance
  • Generates three prediction paths for comprehensive forecasting
  • Produces complete residual tables for each prediction path

Research Methodology

Developmental Research Type II

Combines system development with systematic evaluation, integrating algorithm creation with empirical performance assessment.

Data Collection & Processing

Real parcel shipment data from J&T administrators, with optional log base-10 transformation and 75%/25% train-test split.

Parameter Optimization

Grid Search implementation for automatic optimization of smoothing parameters (α, β, γ) to maximize forecasting accuracy.

Validation Framework

Rigorous testing using MAE, MSE, RMSE, and MAPE metrics with cross-validation techniques on both stationary and non-stationary datasets.

Key Research Findings

Performance Improvements

The GAS framework demonstrated significant improvements over traditional single-algorithm approaches:

3-Path
Forecast Output
75/25
Train-Test Split

Academic Contributions

  • Novel hybrid framework combining SES and HWES methodologies
  • Error-driven adaptation mechanism for dynamic parameter adjustment
  • Multi-path forecasting with weighted SES, weighted HWES, and combined prediction
  • Complete residual computation tables for model interpretability
  • Real-time forecasting simulator for academic and practical applications
Research Results Visualization

Research Contact

Academic Inquiries

For research collaborations, academic discussions, or technical details about the GAS framework.

Authors:

Jherson Aguto & Mark John Panganiban

agutojherson@gmail.com

markjohnpanganiban9@gmail.com

College of Computing Studies, Information and Communication

Isabela State University - Echague Campus | Echague, Isabela

Thesis Advisor: Dr. Ricardo Q. Camungao

Research Implementation

This system serves as the practical implementation component of the thesis for:

  • Course: CS 411 - Thesis Writing 2
  • Program: BS in Computer Science
  • Year: 3rd Year Students Reference
  • Semester: Second Semester 2026-2027

Technical Stack

  • Backend: ASP.NET Core WebAPI
  • Frontend: Vanilla HTML/CSS/JavaScript
  • Database: Microsoft SQL Server
  • Visualization: Chart.js