计算机代写｜COMP7103 Assignment 2

这是一篇关于使用从TEXT0405到TEXT0904中的数据准备一个训练数据集的计算机代写

Question 1 Classification [50 marks]

The Sports articles for objectivity analysis Data Set is the result of extracting features from 1000 sports articles and manually labelled as either “subjective” or “objective” using Amazon Mechanical Turk. A copy of the dataset and the description is also available on Moodle.

Link to the dataset: https://archive.ics.uci.edu/ml/datasets/Sports+articles+for+objectivity+analysis Use Weka and/or any other tools, perform the following.

– Prepare a training dataset using the data from TEXT0405 to TEXT0904. Check that your training dataset consists of an equal number of “subjective” and “objective” records.

– Add one derived attribute “site”, which equals to the domain of the URL. For instance, “site” will be “msn.foxsports.com” for record TEXT0001.

– Find a classification model for the class label “Label” in the dataset. Note that you are expected to try different ideas and refine your model in the data mining process. In particular, you must experiment and document your process of data reduction (see parts a and b below).

Answer the following questions.

a) Briefly describe what you did in constructing the model, including but not limited to data preprocessing, attribute selection, parameter tuning, training and testing data construction, model building, and evaluation, etc. You may omit some of the above-mentioned items if you did not do them. You must clearly describe what your final classification model is, which will be used to answer part c) of the question.

b) Summarize your data reduction process, with justification through model evaluation.

c) Using your final classification model in part a), demonstrate, with explanation, how records

TEXT0001 and TEXT1000 are classified by the model.

Question 2 Association analysis [35 marks]

Table 1 shows a summary of the weather data of Hong Kong from Sep 1, 2022 to Sep 12, 2022.

Attributes Pressure and Temperature are rounded and grouped into consecutive ranges. Rainfall is binarized to “Yes” / “No”.

Date

(Record ID)

Pressure

Temperature Rainfall

1 1008..1012 29..29Yes

2 1003..1007 30..30 No

3 1003..1007 30..30 No

4 1003..1007 31..31 No

5 1003..1007 31..31 No

6 1008..1012 31..31 No

7 1013..1017 28..28 Yes

8 1013..1017 30..30 Yes

9 1013..1017 30..30 No

10 1008..1012 29..29 Yes

11 1008..1012 29..29 No

12 1003..1007 31..31 No

Table 1 Weather data for association rule mining

Perform Quantitative Association Rule (QAR) mining by following the steps below. Show your steps.

a) For attributes Pressure and Temperature, merge the intervals with low support counts with maxsup = 7 (maximum support after merging is at most 7). Consider merging intervals with the lowest support count first.

b) With the merged intervals, transform the dataset in the same way as illustrated in 34 of Chapter 6 notes. For attribute Rainfall, treat Rainfall: Yes and Rainfall: No as two attributes.

c) Use the Apriori algorithm to find all frequent itemsets with a support threshold of 20%.

d) For each frequent itemset, generate rules with confidence of at least 75%.

e) Check if there are rules that could be combined, with confidence of at least 75%.

Question 3 Cluster analysis [15 marks]

*Cluster analysis will be covered near the end of the course

Table 2 shows the location of 6 data objects. Perform hierarchical clustering using Group Average  as the proximity measure. Use Manhattan distance as the distance measure. Show your steps and draw the corresponding dendrogram.

Object

Location

𝑶𝟏 (0, 3)

𝑶𝟐 (9, 33)

𝑶𝟑 (21, 9)

𝑶𝟒 (21, 15)

𝑶𝟓 (33, 15)

𝑶𝟔 (42, 24)

Table 2 Locations of 6 data object