What have we learned?

On this course you have acquired intermediate data science skills and applied them to several different datasets.

Week one

In week one, you started by getting to know the data, which described the artworks held at the Tate Modern Gallery. You selected a subset of the dataset for further analysis, and started by pre-processing the dataset into the appropriate format.

You learned how to visualise the data as part of an exploratory analysis, which enabled you to identify the features to use as the basis for constructing a simple model to predict a portrait or landscape painting on the basis of their dimensions and evaluate the model’s performance. You developed a further model using the words contained in the title text for the artwork and applied the same evaluation.

You were then introduced to complex models, where you were shown how vector mathematics is used in the calculation of distances between locations.

You then covered the topic of classification and clustering where you used the ratios from each model and transformed them into a vector to use as the basis of a more complex model through K-means clustering, which you again evaluated the model and compared the strengths and weaknesses of each of the ratio, title, and K-means models.

You then considered some of the assumptions that should be avoided when discussing how well a model reflects the truth, in other words, the fact that there is bias in both the way we select and filter data, and the techniques we apply to analyse that data.

Week 2

In Week 2 you covered the topic of working with complex data where you learned about distance metrics and classification. You acquired knowledge about the curse of dimensionality: and how to organise data in high-dimensional space. You were then introduced to the importance of feature extraction in data analytics and modelling.

You covered the core topics in machine learning, including techniques, example applications, and the main types of machine learning, including supervised, semi-supervised, unsupervised, and reinforcement learning. You learned about a popular technique using neural networks and how of they are composed of perceptrons, and how they are applied to prediction.

You undertook a practical session on using social media data to make inferences about a population represented by tweets published on Twitter. Here you learned how to pre-process JSON data for visualisation. You discovered how to extract address information from geo-location coordinates stored with the tweets to produce a visualisation based on the mention of certain products in the tweet text.

Given recent changes in how organisations and businesses store, access, and share our data, you were introduced to the ethics of data science, and reviewed a number of guidelines on best practice when applying data analytic techniques to social media data.