7PAM2000 - Applied Data Science 1 Assignment 3 - Clustering and Fitting

Assignment Task

This third assignment will focus on clustering and fitting. This time you are expected to produce a poster which could be used for a poster presentation.

We will again look at exploring public data from the World Bank, and specifically country-by-country indicators related to climate change. You may find additional relevant indicators (e.g. GDP per capita) using the complete list Note that not all countries have entries for the most recent year(s).

Your goal is to

• Find interesting clusters of data. Note that for meaningful clusters it is often a good idea to look at normalised values like GDP per capita, CO ₂ production per head, CO ₂ per $ of GDP or fraction of a You might look at most recent values or compare recent values with, say, values 30 or 40 years ago or use total historic values. Use at least one of the clustering methods from the lecture. Clustering works best when the data are normalised (see Practical 8). Note that you usually want to show the original (not normalised values) to display the clustering results. One way to achieve this is to add the classifications as a new column to the dataframes and use logical slicing. Produce a plot showing cluster membership and cluster centres using pyplot.
• Create simple model(s) fitting data sets with curve_fit. This could be fits of time series, but also, say, one attribute as a function of another. Keep the model simple (e.g., exponential growth, logistic function, low order polynomials). Use the model for predictions, e.g. values in ten or twenty years time including confidence ranges.
• Use the attached function err_ranges to estimate lower and upper limits of the confidence range and produce a plot showing the best fitting function and the con- fidence range.
• You do not need to use the same data sets for clustering and fitting, but one approach could be: find clusters of countries, pick one country from each cluster and compare countries from one cluster and find similarities and differences, compare countries from different clusters or pick a few countries from one clusterand compare with other regions.
• You do not need to focus on CO ₂ or climate The choice of topic is yours.

Content and presentation of the report

Again, you are expected to use your initiative and “tell a story” with the data. You should use appropriate visualisation and provide a text narrative to communicate and explain your findings.

There a some differences between a good report and a good poster. Text in a good poster is shorter. Sometimes use of itemisation or similar could be a good idea. Some information, like comparison between countries, is often better presented in an overview table than in a detailed text.

Not every detail should be explained. A report must contain all necessary information. Readers can ask in a poster session. Still the text should allow a basic understanding without further information. Self-contained graphs are an important tool. Similar to a report good posters often combine information from graphs to draw conclusions or follow up on insights/questions from one graph with another graph.

Your poster should include an introduction and explicit conclusions summarising the important results. Do not include technical information. Your audience is interested in the results not how you got there. Very(!) short remarks on methods used can sometimes we appropriate, but not more.

This 7PAM2000 - Statistics has been solved by our PhD Experts at My Uni Paper.