Collection

2025 Northeast Section Conference

Authors

Shruti Brahma, University of New Haven; Siddhant Alhat Rajendra, University of New Haven; Ardiana Sula, University of New Haven

[19] F. Chollet, “Keras.” https://keras.io, 2015. Accessed: [date you ac-predictive analytics, making it more accessible and actionable cessed]. [20] D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,”for all stakeholders. 2015. [21] S. G. Makridakis, S. C. Wheelwright, and R. J. Hyndman, Forecasting: ACKNOWLEDGMENT Methods and Applications. New York: Wiley, 3rd ed., 1998

Collection

2025 Northeast Section Conference

Authors

Shashi Kiran Chandrappa; Sidike Paheding; Yu Cai Cai

and Llama 3.2,billion parameters, with the 11B and 90B versions optimized allowing for comparative performance evaluation. While bothfor visual recognition and image reasoning, outperforming models classify responses, their output consistency and rea-many existing models on industry benchmarks [16]. soning explanations are analyzed to assess interpretability, ac- 2) Microsoft Phi: Phi-3, developed by Microsoft, is a curacy, and computational efficiency. The classification outputcompact language model designed for efficiency and perfor- is mapped to session security threats such as session fixationmance. The phi-3-mini variant, with 3.8 billion parameters, and hijacking, aiding in real