Critically Evaluate And Implement Appropriate Clustering Algorithms And Interpret: Machine Learning Report, UOL, Ireland

Learning Outcomes Assessed: 

MLO 1 – Critically evaluate and implement appropriate clustering algorithms and interpret and document their results.

MLO 2 –     Perform market basket analysis on representative transactional data. Explore the potential applications of these techniques outside their usual domain.

MLO 3 – Implement text categorization, topic modeling, and document summarization on a range of representative texts. (e.g. Twitter, Facebook)

MLO 4 – Apply modeling to time series data to facilitate business intelligence needs.

Assessment Task

 Students are advised to review and adhere to the submission requirements documented after the assessment task.

Assessment details

 Question 1:

A) Perform the data cleaning and preparation before your visualization and explain your steps. Use at least three visualizations along with specific features to summarise the most important aspects of the data and identify through your visualization why this dataset is suitable for the Machine Learning model in part (b) of Question 1 and provide a logical understanding for that. If you use different visualizations (e.g., statistical graphs, special plots, interactive maps, etc) explain the difference with advantages and disadvantages between

B) Describe the importance of clustering in the domain of unsupervised machine learning. Briefly explain the difference between k-Means and k-Medoid clusterings. Apply both methods to the chosen data set of your choice using machine learning models for the training and testing phases. Evaluate the modeling outcomes and develop at least two illustrations to highlight the similarities.

Question 2:

Discuss the concept and application of Time series analysis using machine learning modeling by providing a real-world data set (It should be different than used in the previous question 1). What is the purpose of The Augmented Dickey-Fuller test in time series?

A) Apply an appropriate ARIMA or Exponential Smoothing model to the chosen data. Check for the model

B) Make one-step-ahead forecasts of the last 10 Determine the forecast errors.

C) Make a time series plot of the data, and further calculate and illustrate the sample autocorrelation and partial autocorrelation (PA). Is there significant autocorrelation in the chosen time series?

D) Create a visualization using Plotly with its functionalities to highlight anomalies/trends/patterns and use the necessary techniques to visualize the density of the data Briefly explain the behavior of your data based on your visualization and use the outcome as a lead to perform a research to complete your explanations.

E) Use a set of subplots to show different times of the series. You must justify the library you selected, reasoning, and segmentation you performed to show valuable insights about the topic.

Question 3:

What is the purpose of a recommendation system for online retail business in machine learning? Briefly compare Content and Collaborative filtering using any dataset of your choice. Train and test machine learning models for the user-user and item-item collaborative filtering. Justify your recommendations for the considered scenario by providing a conceptual insight.

Question 4:

A) Explain the meaning of “correlation” between the variables and use at least three visualizations (Heatmap, Scatter plot, Jointplot, etc) that show correlations in your dataset. Explain the effect of the colors in the graphs and justify your palette choice and

B) Perform Market Basket Analysis on the chosen dataset by using Apriori and FP growth algorithms. Can you express major divergence between these models? Apply unsupervised machine learning models to train and test your data Compare and contrast the machine learning results obtained based on both algorithms.

C) Compare the outcome of your visualization in part a) with the results of the models you performed in part b) of Question Can you associate the results? Briefly explain the connection between the visualization, the data, and the models you performed.