1. Read the question paper carefully and answer all the questions below. Write a document containing your answers, which is saved or exported as a PDF file. Include your name, surname and student number on the first page of your submission file. Any file format other than a PDF file is unacceptable for submission, and will not be assessed.
2. You may consult any sources you wish, including any additional articles you can find. Include accurate and complete references for all the external sources you refer to. Do not include text taken directly from any sources, even if you quote it correctly. Excessive direct quoting will be penalised. Paraphrase your answers, and give your own opinion. Note that not all sources are reliable or correct, even if they are published in conference proceedings or a journal (in other words, be critical of every source you consult).
You may not consult with any other individual with regard to any part of this Assignment. Any such external consultation is considered plagiarism, and will not be tolerated.
3. Note that the above point does not imply that everything you mention must come from a source. Many of the questions test your interpretation and understanding of concepts covered in the course content, or related to the course content. Well reasoned arguments will receive credit. Conversely, unsubstantiated or dubious observations from cited sources will not receive credit.
4. By submitting your answers you implicitly agree to the plagiarism policy of the Computer Science De- partment at the University . This policy is repeated below, for your convenience. If you are found guilty of any plagiarism, disciplinary action will be taken. This may include suspension of your studies at the University of Pretoria. Note that your submission will be tested for plagiarism using the Turnitin system. Your submission will not be permanently uploaded into the Turnitin database.
5. Write clearly and concisely. The duration of the Assignment is intended to allow you to carefully consider your answers, and does not imply anything about the length of your submission. If you are asked to discuss multiple points, use bulleted lists to clearly organise your answer. Be sure that the numbering of your answers is correct. If a question asks you to provide a certain number of items (e.g. a certain number of advantages or disadvantages), you will be penalised if you provide more than the required number.
The Department of Computer Science considers plagiarism to be a serious offence. Disciplinary action will be taken against students who commit plagiarism. Plagiarism includes copying someone else’s work without consent, copying a friend’s work (even with consent) and copying material from the Internet. Copying will not be tolerated in this module.
1. A spatial database can be represented in either a raster or vector format. Identify one general complexity associated with performing data mining on a raster format spatial database, which is not associated with performing data mining on a vector format spatial database.
2. Decision trees and production rules can be used to classify data samples. Describe only one general drawback associated with the use of traditional decision trees for classification, which is not associated with the use of traditional production rules for classification. Explain both why the drawback is associated with decision trees, and why it is not associated with production rules.
3. One complexity associated with incremental mining is that the model created by the mining process should progressively forget stale and irrelevant patterns. Identify only one additional complexity associated with incremental mining.
4. Suggest only one possible drawback that could be associated with the re-integration of extracted knowledge into the data source, as part of the KDD (knowledge discovery in databases) process.
1. One simple approach to replacing a missing value for a continuous attribute in a data sample is:
• Determine the class of the data sample with the missing value.
• Compute the mean of the missing attribute value over all data samples that also have this class.
• Replace the missing attribute value with this mean.
Identify only one situation in which this approach to missing value replacement would be inappropriate, and justify your answer.
2. One problem associated with data integration is the entity identification problem, in which difficulties may arise when trying to identify which attributes are equivalent to one another in two or more data sets that need to be combined. Metadata describing the nature of attributes in the various data sets may assist in the identification process. Imagine that two data sets need to be integrated and the entity identification problem arises, but no metadata is available. Suggest only one approach that could be used to resolve the entity identification problem by using only the data sets themselves (i.e. with no external assistance).
3. Feature construction refers to the construction of new attributes from existing attributes in a data set. Imagine that a data set contains general demographic data for the population of an entire country, and has an attribute named pregnancies, which indicates how many times an individual has been pregnant. Furthermore, a data analyst decides to construct a new categorical attribute named sex. The analyst writes a script that sets the sex attribute value for an individual to female when the number of pregnancies for the individual is one or more, and sets the value to male when the number of pregnancies is zero. Suggest only two general drawbacks that are associated with this approach.
1. A very commonly used visual dimension in data visualisations is colour. Identify only two drawbacks that are specifically associated with using colour as a visual dimension.
2. Exploratory data analysis focuses on data visualisation. Briefly explain how a scientific visualisation could be converted into a data visualisation for exploratory data analysis purposes.
3. An area chart is a data visualisation technique. Identify only one drawback associated with area charts.
1. Identify only one complexity associated with divisive hierarchical clustering algorithms, which is not as- sociated with agglomerative hierarchical clustering algorithms.
2. Consider the k-means and k-medoids clustering algorithms. Unseen data samples are data samples that were not used by the clustering algorithm to construct clusters, but still have the same underlying characteristics as the data samples used by the algorithm. We say that a clustering algorithm generalises well if the clusters produced by the algorithm correctly represent any unseen data samples. Identify whether the k-means algorithm or the k-medoids algorithm is likely to generalise better, and justify your answer.
3. Consider the cluster quantisation error quality measure. Identify only one situation in which the cluster quantisation error will produce an inaccurate assessment of cluster quality, and justify your answer.
1. Most decision tree building algorithms use a greedy approach. This means that once an attribute test has been chosen, the algorithm cannot backtrack to change the attribute test. Answer the following questions:
(a) Briefly describe only one advantage of this greedy approach.
(b) Briefly describe only one disadvantage of this greedy approach.
2. One advantage of decision tree pruning is that it can result in more compact trees that are more understand- able to human data analysts. Identify another possible advantage associated with decision tree pruning, and briefly explain why this advantage exists.
3. The C4.5 algorithm can convert decision trees into rule sets. One of the final steps in this conversion process groups rules into groups (where each group contains all the rules predicting the same class), and orders these groups in ascending order of the number of false positives generated by the group. Briefly explain why this ordering is sensible.
1. The AQR algorithm repeatedly selects a positive example called a seed. A seed is used to generate complexes by repeatedly choosing a negative example (i.e. an example that is incorrectly covered by a complex, because its class does not match the class of the seed). The algorithm then specialises the complexes to exclude the negative example. AQR chooses the negative example that is closest to the seed. Briefly explain why this choice of a negative example is a good strategy.
2. When the CN2 algorithm specialises complexes using a continuous attribute, the attribute’s range is divided into equal sized sub-ranges. The size of sub-ranges is a single user-defined parameter for the algorithm. A set of selectors is then constructed, where each selector tests whether the attribute’s value either exceeds or does not exceed a value at a sub-range boundary. Identify only two drawbacks to this approach.
3. The CN2 algorithm prunes the worst complexes from the STAR set at the end of each series of complex specialisations. The amount of pruning that takes place is controlled by a user-specified limit on the size of the STAR set. A smaller size limit will result in more pruning, while a larger size limit will result in less pruning. Briefly explain how the CN2 algorithm’s search for a best complex will be affected by choosing a smaller size limit on the STAR set.
1. The learning rate and neighbourhood radius of a self-organising map should both decrease as training continues. Briefly explain what the result will be if these values do not decrease in this way.
2. Identify only one disadvantage associated with weight-centric neuron labelling for self-organising maps.
3. In the context of self-organising maps, unsupervised example-based neuron labelling can be seen as an unsupervised version of example-centric neuron labelling. Identify two reasons that unsupervised example- based neuron labelling is much more complex for a human data analyst to perform than example-centric neuron labelling.
4. Briefly describe only one drawback associated with the SIG* algorithm.
CS 340 Milestone One Guidelines and Rubric Overview: For this assignment, you will implement the fundamental operations of create, read, update,
Retail Transaction Programming Project Project Requirements: Develop a program to emulate a purchase transaction at a retail store. This
7COM1028 Secure Systems Programming Referral Coursework: Secure
Create a GUI program that:Accepts the following from a user:Item NameItem QuantityItem PriceAllows the user to create a file to store the sales receip
CS 340 Final Project Guidelines and Rubric Overview The final project will encompass developing a web service using a software stack and impleme
