Midterm Presentation Guide

Goal and Expected Quality

This guide helps your team present the Midterm Project Baseline with the structure and clarity expected in a scientific conference talk.

• Keep a clear storyline: problem → method → results → limitation/next steps.
• Use clean visualizations and concise slide text (avoid long paragraphs).
• Every claim should be supported by data, plots, or metric tables.

Recommended Slide Organization

1) Problem Definition

• Background: Introduce domain context, why this time-series problem matters, and practical relevance.
• Macro / societal / practical framing (required): Do not define the problem only in theoretical terms. Explain impact at three levels:
  • Macro level: Economic/system-level importance (cost, productivity, planning, risk management).
  • Societal level: Effects on people and communities (health outcomes, fairness, safety, service quality).
  • Practical level: Operational decision impact (how your forecast/classification changes actions in real workflows).
• Limitation / Challenge: Describe core difficulties (noise, non-stationarity, seasonality, missing data, class imbalance, etc.).
• Leading to motivation: Explain why these limitations motivate your proposed approach.

Example problem framing (recommended slide pattern):

• Healthcare calls forecasting:
  • Macro: Better staffing reduces overtime and emergency overload.
  • Societal: Faster response improves patient safety and trust in public health services.
  • Practical: Daily call-volume forecasts guide shift allocation and escalation planning.
• Electricity load forecasting:
  • Macro: Accurate load planning reduces balancing costs and supports grid stability.
  • Societal: Fewer outages and more reliable power for households and hospitals.
  • Practical: Hour-ahead forecasts optimize dispatch and peak-load mitigation.

2) Method

• Dataset: Source, size, sampling frequency, target variable, and train/validation/test split.
• Analysis of dataset: Key EDA findings (trend/seasonality/distribution/anomalies/correlation), with visuals.
• Proposed model: Pipeline overview, features/preprocessing, model architecture, and training strategy.

3) Results

• Experiment setting: Evaluation metrics and baseline/comparison methods.
• Experiment results and discussion: Tables/plots + interpretation (what improved and why).
• Limitation of current methods and next plan: Honest limitations, failure cases, and concrete next-phase plan.

Conference-Style Presentation Checklist

• Start with a 1-slide problem statement and 1-slide contributions.
• Use one visual message per slide (avoid overloaded charts).
• Label all figures clearly (units, axis names, legend, metric definition).
• Keep methods reproducible: mention important hyperparameters and validation setup.
• Mandatory: Include a source code link directly in the slides (GitHub repository, Kaggle notebook, or another accessible version) so reviewers can reproduce your pipeline.
• Include at least one ablation/comparison insight, not only final numbers.
• End with takeaway bullets: what works, what fails, what’s next.

Suggested Timing (5–7 minutes)

• 1.0–1.5 min: Problem Definition + Motivation
• 2.0–2.5 min: Method (Dataset, EDA, Proposed Model)
• 1.5–2.0 min: Results + Discussion
• 0.5–1.0 min: Limitation + Next Plan + Closing