ENG335: Group-based Assignment SUSS- Machine Learning, Singapore

Question 1

(a) Read about GPT and the malicious clones of GPT. Summarize in a paragraph not more than 10 lines on what you have learnt and understood.

(b) How is ChatGPT able to reply to questions related to data after Sep 2021? Is ChatGPT a search engine?

(c) Why trained models are given as open source by companies? Cite THREE (3) reasons.

(d) Read about Google’s Gemini. Discuss the differences between Gemini and ChatGPT (Limit your answer to one A4 page)

Question 2

The dataset is available in the below link.
https://www.kaggle.com/datasets/abrambeyer/openintro-possum

(a) Perform exploratory data analysis and understand the parameters. Encode the ‘Pop’ and ‘sex’ features. Use 4 for the ‘NA’ values in ‘age’ and 68 for ‘NA’ values in ‘footlgth’.

(b) The objective is to estimate the possum’s head length. Estimate the possum’s head length using the best FOUR (4) features and present the estimation model.

(c) Assess the performance of the linear regressor by getting the relevant performance metrics. You need to provide any THREE (3) metrics and explain the importance of these metrics. Use at most 15% of the dataset for testing.

Question 3

Download the “Palmer Penguins” dataset from Kaggle
(https://www.kaggle.com/datasets/satyajeetrai/palmer-penguins-dataset-for-eda ).

(a) Perform exploratory data analysis and understand the dataset. Drop the ‘year’, ‘island’ and ‘species’ features. Use Python code to discard the data records with missing values. Implement a suitable algorithm from what you have learned in the class for identifying the gender of penguins.

(b) Construct a Naïve Bayes algorithm for the above dataset.

(c) Use at most 15% of the dataset for testing. Compare the performance metrics of the algorithm in Question 3(a) and Naïve Bayes classifier. Does the scaling of the parameters have any impact on the performance (Justify your answer)?

Question 4

Use the Kaggle dataset available in https://www.kaggle.com/datasets/cvergnolle/gold-priceand-relevant-metrics. Use the data records only from 14 Jan 2022 to 24 July 2024, inclusive of
both dates. The objective is to check if Crude price has any correlation with the other features in the dataset and propose a suitable model to predict the Crude price. Use 20% of the dataset for testing. Your model should have a R2_score of at least 0.70.