ACCT 331 WEEK 9

ACCT 331: Introduction to Applied Artificial Intelligence

  • Course details:

    • Location: Schreiber Hall #302

    • Schedule: Tuesday and Thursday, 1:00-2:15 PM

    • Course Number: 10304

News You Can Use

Walmart Partnership with OpenAI

  • November 14, 2025

  • Overview of the partnership:

    • Walmart and Sam's Club utilizing AI to enhance shopping experiences.

    • Collaboration allows customers to complete purchases within ChatGPT.

    • Transition from reactive to proactive retail experiences through AI.

  • Key insights:

    • Customers' needs anticipated by AI systems.

    • AI influence on product catalogs, customer care, and internal training.

JPMorgan Chase 2026 AI Research Associate Program

  • Job information:

    • Role: AI Research Associate Intern, New York, NY

    • Posting date: September 18, 2025

    • Salary range: $135,000-$155,000.

  • Internship goals:

    • Advance AI research related to machine learning and cryptography.

    • Collaborate with experts across multiple fields and academic institutions.

Internship Opportunities

  • Data Science Internship at IBM Quantum

    • Available for summer; contact: Daniella Pombo

  • Internship at CME Group

    • Seeking interns with strong CS/AI/ML backgrounds.

    • Contact: Vijay Pillai (vijay.pillai@cmegroup.com)

AI Landscape Insights

Byron Deeter's Perspective on AI and Venture Capital

  • Statement: AI as the “technology opportunity of our lifetimes.”

  • Predictions and expectations:

    • AI surpasses the cloud computing wave in potential.

    • Future applications will focus on user time-saving and productivity enhancement.

    • Potential disruption to established tech giants.

Data Science Trends

Current Landscape

  • Data management and analytics retain a vital role in tech investments.

  • Demand for specialized data roles over generalist ones.

  • Companies focus on operationalizing AI to achieve measurable outcomes.

Cisco's AI Research Findings

  • AI Adoption Statistics:

    • AI-ready companies outperform peers; 4x higher production conversion.

    • Concerns regarding barriers to efficient AI implementation such as GPU capacity.

Course Schedule and Topics

Exam and Assignment Dates

  • Important Dates:

    • SKP Release: October 23, 2025;

    • Assignment Due: October 30, 2025.

Content Overview (Week 9)

  • Topics:

    • Model evaluation for classification.

    • Model selection.

    • Model interpretability and fairness.

    • Machine learning use cases.

Measuring Error in Regression Models

Key Metrics and Concepts

  • Goal: Assess accuracy of regression models using various metrics:

    • R² Statistic (R²) and Adjusted R²

    • Mean Absolute Error (MAE)

    • Mean Squared Error (MSE)

    • Root Mean Squared Error (RMSE)

Error Function

  • Definition: Analyzes performance by comparing predictions to actual data points.

  • Minimal values indicate good predictions.

Regression Model Assumptions

  • Model depicted as: Y=f(X)+ϵY = f(X) + ϵ

  • Where ϵ is a mean-zero error term, emphasizing the challenges in perfect prediction.

Measuring Accuracy for Classification Models

Fundamentals of Classification Model Evaluation

  • Need for accuracy quantification for:

    • Model competition

    • Feature selection

    • Performance reporting

Common Classification Metrics

  • Important definitions and metrics:

    • Accuracy: extAccuracy=racTP+TNTP+TN+FP+FNext{Accuracy} = rac{TP + TN}{TP + TN + FP + FN}

      • Need balancing with other metrics in imbalanced datasets.

    • Precision (Positive Predictive Value): extPrecision=racTPTP+FPext{Precision} = rac{TP}{TP + FP}

      • Minimize false positives.

    • Recall (Sensitivity): extRecall=racTPTP+FNext{Recall} = rac{TP}{TP + FN}

      • Important in critical detection tasks to minimize false negatives.

    • F1-Score: Harmonic mean of precision and recall, given by:
      F1=rac2imes(PrecisionimesRecall)Precision+RecallF1 = rac{2 imes (Precision imes Recall)}{Precision + Recall}

    • ROC Curve: Visual representation of true positive rate vs. false positive rate across thresholds.

Confusion Matrix Definition

  • Tool for evaluating model performance, showing breakdown of:

    • True Positives (TP)

    • False Positives (FP)

    • True Negatives (TN)

    • False Negatives (FN)

Misleading Aspects of Accuracy

  • Issues arise in imbalanced datasets.

  • Example scenario involving classification for rare diseases showcases the potential pitfalls of high accuracy indicators without enabling practical usefulness.

Conclusion and Key Metrics

Summary of Evaluation Metrics:

  • Accuracy: Overall correctness without class weighting.

  • Precision: Focus on reliability of positive predictions.

  • Recall: Assess ability to capture actual positives.

  • F1-Score: Balance between precision and recall.

Importance of Understanding Error Types

  • Distinguish between false positives and false negatives in critical real-world scenarios such as medical diagnoses and criminal justice.

Model Selection and Interpretability

Definition of Model Selection

  • Process of choosing best-performing models for specific tasks based on defined metrics.

  • Includes hyperparameter tuning and selection across various algorithms.

Criteria for Consideration

  • Model training duration, interpretability, and overall complexity.

  • Trade-off between model accuracy and explanation ease may influence final selection.

Need for Interpretability

  • Essential in high-stakes scenarios, where understanding model reasoning is necessary for effective application.

  • Complex models (black boxes) vs easy-to-interpret models.

Techniques for Enhancing Interpretability

  • Feature importance scores

  • Model summaries

  • Examination of prediction contributions

Ethical Considerations in AI Models

Fairness in AI Systems

  • Definition and importance of developing unbiased AI systems.

  • Typical causes of model bias, including data quality and feature selection.

Regulatory Landscape

  • Recent regulations emphasize data protection, algorithmic transparency, AI ethical use, and sector-specific legislation.

Business Issues and Case Studies

Model Applications to Reduce Stock-Outs

  • Overview of modeling approaches to manage inventory through demand prediction.

Statistical Insights on Customer Behavior

  • Analysis of basket composition to improve customer shopping experiences and reduce stock-outs effectively.

Revenue Impact Analysis from Model Implementation

  • Documented ROI and successful application of predictive models in enhancing customer satisfaction and inventory management.