Machine Learning Portfolio Report

Return to Home

Word Count (excluding references): 2303

Collaborative Discussion 1: The 4th Industrial Revolution

Recent literature describes a shift from the efficiency-driven and automated focus of Industry 4.0 to the more human-centric and ethical principles that define Industry 5.0. The analysis of the Marks & Spencer cyberattack in the UK demonstrates how rapid adoption of centralized digital systems can create significant vulnerabilities and lead to large-scale disruption (Gallagher, 2025; Ravikumar & Young, 2025). Such incidents not only impact operations and finances but also diminish customer trust (Mittelstadt & Floridi, 2017). These events highlight the need to prioritize risk management, transparency, and human oversight alongside technological advancement (Heuser & Wang, 2021; Metcalf, 2024).

Industry 5.0 is presented as a rethinking of the relationship between technology and people, rather than a simple upgrade. The expectation is that future systems will be adaptable, ethical, and resilient to cyber threats. This includes integrating explainable, AI-driven security mechanisms and involving employees directly in risk detection and response, especially as IoT devices and human-AI interfaces expand the system's complexity and vulnerability (Santos et al., 2024; Rožanec et al., 2023).

Parallel developments are taking place in software testing and quality assurance, where artificial intelligence, machine learning, and big data analytics have increased efficiency and accuracy (Antony et al., 2022a; Alzahrani et al., 2021). However, these technologies also bring new ethical and security challenges. Frameworks such as DevSecOps and privacy-by-design are now seen as important for balancing innovation with responsible governance (Mack, 2023; Schwab, 2016).

In summary, current thinking suggests that technological progress should align with human values, ethics, and a sustained focus on security and public trust. Achieving the benefits of Industry 4.0 and 5.0 depends on maintaining this balance.

References

Collaborative Discussion 2: Legal and Ethical Views on ANN Applications

Recent discussions in the literature and among practitioners highlight both the capabilities and limitations of AI-powered language models such as GPT-3. These tools can generate fluent, human-like text quickly and have practical value for administrative tasks, report drafting, and summarization, offering potential gains in efficiency and reduced workload (Zhou et al., 2020). In professional settings, these efficiencies can be valuable, and many workplaces may benefit from the adoption of such systems.

However, the use of AI writing tools in creative and educational contexts raises important concerns. Scholars note that language models do not possess genuine understanding; they operate by predicting the next word in a sequence based on vast training data, rather than any semantic or contextual grasp of meaning (Bender et al., 2021). This has implications for creative writing, where the distinction between authentic artistic expression and algorithmically generated text can be blurred, raising questions about the true value of AI-generated literature (Hutson, 2021).

Ethical issues are also prominent. Language models trained on internet-scale datasets can absorb and reproduce biases, stereotypes, and misinformation present in their source material, making them a potential vector for harm if used without careful oversight (Weidinger et al., 2021). There are growing calls for educational institutions and organizations to provide AI literacy training, implement clear usage policies, and ensure that human oversight is central to all creative or factual outputs generated by AI systems. Concerns over originality, consent, and data ownership in the creative industries further underscore the need for regulation and thoughtful integration of these technologies (Higgs & Stornaiuolo, 2024).

Other voices argue that with the right frameworks and human supervision, AI can enhance creativity, provide new perspectives, and support professional writing, provided its limitations are understood and ethical safeguards are in place. Ultimately, the consensus is that while AI writing tools can support efficiency and creativity, their use must be balanced by transparency, responsibility, and ongoing critical evaluation.

References

e-Portfolio Summary: Applied Learning in Machine Learning

The e-Portfolio activities provided valuable, hands-on experience with core machine learning concepts and methods. By working through a range of Jupyter Notebooks, I developed a clearer understanding of how theoretical principles are applied in practice and how careful model selection and performance evaluation affect results.

The first set of exercises focused on foundational statistics, including covariance and Pearson correlation. Calculating and interpreting these measures helped me understand how variables relate and which ones are best suited for prediction. The linear regression notebooks guided me through building models, interpreting their coefficients, and using metrics such as R² and mean squared error to evaluate performance. When I moved to multiple regression, I observed the impact of adding more predictors, including issues like noise and multicollinearity. Polynomial regression offered a way to capture nonlinear trends but also highlighted the risk of creating models that fit the training data too closely.

The Jaccard coefficient exercise was useful for measuring similarity in categorical data. This metric proved especially practical for identifying clusters in medical test results and demonstrated the importance of selecting features thoughtfully.

Exploring neural networks, starting from simple perceptrons and advancing to multi-layer architectures, showed how these models capture complex patterns. The exercises also demonstrated that tuning hyperparameters is necessary to avoid poor generalization. Visualizations and direct coding practice made the concepts more accessible.

The CNN object recognition notebook illustrated how convolutional layers process images for classification tasks. By experimenting with model predictions, I observed both strengths and weaknesses of the approach. Reading about the ethical and social implications, such as bias and privacy concerns, helped frame the technology's impact on society and the need for transparency.

Model performance measurement activities emphasized how parameter tuning influences accuracy and reliability. Adjusting factors like tree depth or regularization demonstrated that careful evaluation is essential for trustworthy models.

Finally, the gradient descent exercise showed how iterative optimization works in machine learning. Varying learning rates and iteration numbers provided insight into finding a balance between speed and stability. The Mayo (2017) reading reinforced the importance of both mathematical and practical considerations when applying optimization techniques.

Overall, these e-Portfolio tasks built a solid foundation in machine learning. The combination of hands-on work and critical reflection improved my ability to analyze data, evaluate models, and consider the ethical aspects of deploying machine learning in real-world situations.

Reflective Analysis: Airbnb Advanced Analytics Project

As part of the Advanced ML track, our team addressed the question: How can Airbnb use anomaly detection and trend identification to improve strategic decisions about pricing across New York City? We applied self-supervised learning for anomaly detection and leveraged Neural Architecture Search (NAS) to optimize our models.

Collaborating virtually brought the usual teamwork challenges, such as aligning on project direction and ensuring clear communication. Early on, we debated how to define an “anomaly” and spent time standardizing our approach through exploratory data analysis and team discussion. Explaining complex concepts like self-supervised learning and NAS required breaking down ideas into accessible terms for all team members, which helped deepen my own understanding.

Technically, the project presented significant challenges. The dataset required extensive cleaning and feature engineering to support our machine learning approach. Defining self-supervised tasks and running NAS involved experimentation and careful resource management. Despite these hurdles, our models successfully detected pricing outliers and emerging trends, including shifts tied to holidays and sudden demand changes in certain neighbourhoods. We worked to present these insights to Airbnb’s executives in a clear, business-focused manner, emphasizing the commercial value and strategic implications of our findings.

Throughout, we stayed mindful of ethical and professional issues. We considered the risk of misinterpreting anomalies and were careful to maintain data privacy and fairness. Peer feedback and group reflection helped us improve both technically and as a team.

Overall, this project strengthened my skills in advanced analytics, communication, and ethical awareness. I learned how to critically apply cutting-edge ML methods to real business problems and how to convey technical findings to diverse audiences. These experiences will be valuable for future roles in analytics and data science.

Ultimately, the project provided a comprehensive view of how advanced machine learning techniques can be leveraged for real-world business analytics. I learned to critically appraise both the strengths and the limitations of self-supervised approaches, and to communicate their value and risks to both technical and non-technical audiences. By the end of the project, I felt more confident in my ability to work as part of a diverse analytics team, to handle uncertainty in machine learning workflows, and to consider the broader legal, social, and ethical implications of deploying advanced ML in a commercial context. This experience will inform my approach to future analytics projects, particularly those that require the integration of technical rigor, business insight, and ethical awareness.

Project Proposal (Unit 6) vs. Final Project (Unit 11) Comparison

For my final project, I used advanced machine learning techniques on the CIFAR-10 image dataset, focusing on self-supervised learning with SimCLR, neural architecture search (NAS), and transfer learning. This approach required detailed data preparation, including splitting data for unbiased evaluation and applying real-time augmentation to improve model robustness. In contrast, the Airbnb project relied mainly on traditional supervised learning, classical regression, and clustering, with most of the work centered on tabular data cleaning and feature engineering.

One key lesson was the value of adaptability. While I aimed to use full self-supervised learning and NAS, computational constraints in my environment made it necessary to pivot to transfer learning with MobileNetV2. This experience was similar to the Airbnb project, where practical limitations also influenced our analytic approach.

Evaluation strategies differed between the projects. In the Airbnb work, I focused on business-oriented metrics like R² and visualizations for actionable insights. For CIFAR-10, I relied on accuracy, precision, recall, and confusion matrices to understand performance across classes. Data augmentation improved model recall in some categories, even if overall accuracy stayed the same.

Both projects emphasized the importance of ethical and professional standards. Careful validation, clear reporting, and an awareness of bias were necessary throughout. Communicating not only results but also limitations was critical for transparency.

Overall, the final project allowed me to deepen my knowledge of advanced ML, automation, and image analysis, while reinforcing practical skills like problem-solving and critical evaluation. Compared to the Airbnb analysis, which prioritized business application and interpretability, the final project highlighted the complexity and adaptability needed for state-of-the-art machine learning. Both approaches strengthened my ability to apply analytics to real-world challenges.

Unit Reflection

Knowledge of Machine Learning Algorithms

This module began with foundational topics such as correlation, regression, and the mechanics of model selection. By applying these concepts in Jupyter Notebooks, I strengthened my skills in interpreting variable relationships, evaluating model fit, and recognizing the limitations and potential pitfalls of regression analysis, including issues like overfitting and multicollinearity.

Progressing to neural networks and deep learning, I learned how these architectures can capture more complex relationships in data compared to traditional models. Implementing simple perceptrons and multi-layer neural networks offered insight into supervised learning workflows, model training, and the importance of hyperparameter tuning. The exploration of convolutional neural networks for image recognition allowed me to observe the unique challenges and strengths of advanced models. These practical activities solidified my understanding of how different algorithms are suited to particular tasks and data types.

In the final project, I applied more advanced techniques such as self-supervised learning and neural architecture search. These methods introduced additional layers of complexity, especially in terms of data preparation and model selection. The experience emphasized the need for adaptability, as practical limitations often require researchers to adjust their approach. Using advanced models also required careful evaluation with appropriate metrics, such as precision, recall, and confusion matrices, to ensure that performance assessments were robust and meaningful.

Teamwork and Collaboration

Teamwork played an important role in this module. Participating in a virtual team required clear communication, flexibility, and mutual respect. Team members collaborated to divide project tasks, share ideas, and provide feedback throughout the analytic and reporting phases. Regular meetings and shared documentation supported effective collaboration and project management. While there were occasional challenges in aligning schedules or perspectives, a collective focus on project goals and professional conduct helped us navigate these issues.

Through team activities, I learned to value diverse viewpoints and approaches to problem-solving. The exchange of feedback and ideas within the group contributed to more thorough analyses and encouraged critical reflection. The experience of working collaboratively in a virtual environment improved my interpersonal and communication skills and provided perspective on how teamwork can support both individual and collective learning.

Professional and Personal Development

The learning activities in this module reinforced several key professional skills. Time management was crucial for balancing individual assignments, team responsibilities, and deadlines. Careful planning and progress tracking supported steady workflow and helped address unforeseen obstacles efficiently. The need to communicate technical findings to both technical and non-technical audiences improved my clarity and conciseness in written and oral communication.

The module also emphasized legal, ethical, and social considerations in machine learning. Collaborative discussions explored topics such as the shift from Industry 4.0 to Industry 5.0, the risks associated with AI-generated content, and the importance of transparency, privacy, and fairness in AI deployment. These discussions highlighted the responsibilities of computing professionals beyond technical problem-solving, particularly regarding the societal impacts of technology.

The hands-on portfolio activities, including work on regression, classification, and model optimization, improved my programming skills, data literacy, and critical thinking. Evaluating model performance, addressing data quality issues, and reflecting on the limitations and ethical considerations of each project deepened my understanding of applied analytics.

Conclusion

In summary, this module provided a comprehensive education in both the technical and professional aspects of machine learning. I gained practical experience with a wide range of algorithms and learned to apply them thoughtfully to different types of data and analytic challenges. Teamwork contributed to my growth in communication and collaboration, while independent work allowed for self-directed learning and critical reflection. The emphasis on ethical and societal considerations reinforced the importance of responsible practice in machine learning and computing. I am now better prepared to contribute meaningfully to analytics projects and to approach future challenges with an informed and reflective mindset.