锐客网

  1. 首页 > 睿知 > it技术 > >

SIT742 数据分析

算法

【SIT742 数据分析】SIT742 (Modern Data Science)
Full Marks: 40
Assessment Task 02
2021 Trimester 1, Due: 8:00pm AEST, 22/05/2021
Students with difficulty in meeting the deadline because of illness, etc. must apply for an
assignment extension in CloudDeakin no later than 12:00pm on 21/05/2021 (Friday).
? This is a group work for group with up to 3 members. If you choose to work on it
individually, please seek approval from unit chair via email.
? There are folders for this task on CloudDeakin, please enrol into the group (2 or 3
members) before 15/05/2021 (12:00am):
2021 Assessment Task 2 (1-member Group) for students with approval of
working alone; Approval required;
2021 Assessment Task 2 (2-member Group) for groups of 2 members; Selfenrollment
required;
2021 Assessment Task 2 (3-member Group) for groups of 3 members; Selfenrollment
required.
? Please form the group first, and then self-enrol into the appropriate group before
15/05/2021 (12:00am).
Instructions
Six files are provided for this assessment task:
HTWebLog_p1.zip The compressed zip file is for Part I of this assessment task, and it is a sample of Hotel
TULIP Web log dataset, which contains the web access log information from 11/2006 to 02/2007. 1
.
Citation2003-2021.csv This CSV file is for Part II of this assessment task, and the file structure is provided.
Search-results.csv This CSV file is for Part II of this assessment task, and the file structure is provided.
SIT742Task2.ipynb This is the notebook file for the Python code in ipynb, and the latest notebook is also
released in SIT742Task2.ipynb.
Web log This code snippet contains all the coding requirements and also hints for Part I of this
assessment task.
Predictive Aanalysis This code snippet contains all the coding requirements and also hints is for
Part II of this assessment task.
You will need to complete the code in the notebook and make it run-able. The results of running the
notebook will help you to develop your report, as well as generate the required files: Citation2003-
2021.csv and Search-results.csv.
SIT742Task2-Report-Template.docx This is the Word template for your report SIT742Task2-Report.pdf.
What to Submit?
You are required to submit the following completed files to the corresponding Assignment (Dropbox) in
CloudDeakin:
SIT742Task2.ipynb The completed notebook with all the run-able code on all requirements.
SIT742Task2-Report.pdf Your report for the both Part I and Part II of this assessment task.
1This file is exclusively for SIT742 educational purpose only. You are not allowed to further distribute it.
Page 1 of 6
SIT742 (Modern Data Science)
Full Marks: 40
Assessment Task 02
2021 Trimester 1, Due: 8:00pm AEST, 22/05/2021
Citation2003-2021.csv The completed citation information as CSV file, sorted by year.
Search-results.csv The completed parameter grid search result as CSV file.
Part I
Data Analytic — Web Log Data (20 marks)
Here is the hypothetical background:
Hotel TULIP (a hypothetical organisation) is a five star hotel that locates in Australia. It is a
very special hotel with an equally special purpose: Not only does it embody all the creative energy
and spirit of TULIP-Lab, it’s a “learning environment” on which the tourism and hospitality
students are trained for future hoteliers.
In the past two decades, the Web server of Hotel TULIP has logged all the web traffic to the
hotel website, and stored large amount of data related to the use of various web pages. The hotel’s
CIO, Dr Bear Guts (not Bill Gates!), believes that those log files are great resources to help their
Information Technology Division improve their potential customers’ online experience, and help
their Market Promotion Division to identify potential customers and their behaviour patterns.
Hence, Hotel TULIP would like you Group-SIT742 (a hypothetical data analytics group with up
to 3 data analysers) to analyse web log files and discover user accessing patterns of different web
pages.
The Web server is using Microsoft Internet Information Service (IIS), and the Web log format
can be found at: https://msdn.microsoft.com/en...(v=vs.90).aspx
Task Description
This task requires you to develop a data analysis report for the provided Hotel TULIP Web logs.
Without exploration or further analysis, ‘raw’ Web log data hardly reveals any insightful information.
In this part, you are required to complete the Python code snippets to generate suitable numeric and visual
description in the Hotel TULIP Web log dataset based on the detailed requirements in SIT742Task2.ipynb,
and develop the report SIT742Task2-Report.pdf to summarise the data analytic results. The detailed
requirements can also be found in the notebook SIT742Task2.ipynb, here we summarise them as follows:
1 Data ETL (4 marks)
1.1 Load Data (2 marks)
Load data from files. In order to reduce the processing time, we will remove missing values, and select 30%
of total data for the following tasks.
Code ? Remove missing values. For the columns, if the column is with 15% NAs, you need to remove
that column. Then, for the rows, if there are any NAs in that row, you need to remove that row
(requests)
? Randomly select 30% of the total data into a new dataframe weblog_df.
Report ? Please show the number of requests in weblog_df.
1.2 Feature Selection (2 marks)
Code Select ’cs_method’, ’cs_ip’, ’cs_uri_stem’, ’cs(Referer)’ as features and ’sc_status’ as the class
label into a new dataframe ml_df for following Machine Learning Tasks.
Page 2 of 6
SIT742 (Modern Data Science)
Full Marks: 40
Assessment Task 02
2021 Trimester 1, Due: 8:00pm AEST, 22/05/2021
Report ? Data Description of ml_df.
? Show the top 5 rows of ml_df.
2 Unsupervised learning (4 marks)
You are required to complete this part using sklearn.
Code ? Perform unsupervised learning on ml_df with K Means.
Report ? Visualization of ‘KMeans’ performance using the elbow plot , with a varying K from 2 to 10.
? What is the best K for this dataset?
3 Supervised learning (8 marks)
You are required to complete this part using PySpark packages.
3.1 Data Preparation
Prepare the data for supervised learning by completing the code.
3.2 Logistic Regression (4 marks)
Code ? Perform supervised learning on ml_df with Logistic Regression.
? Evaluate the classification performance using confusion matrix including TP, TN, FP, FN;
? Evaluate the classification performance using Precision, Recall and F1 score.
Report ? Show the classification result using confusion matrix.
? Evaluate the classification performance using confusion matrix including TP, TN, FP, FN,
? Evaluate the classification performance using Precision, Recall and F1 score.
3.3 K-fold Cross-Validation (4 marks)
You are required to use K-fold cross validation to find the best hyper-parameter set where K = 2.
Code ? Implement K-fold cross validation for three (any three) classification models.
Report ? Your code design and running results.
? Your findings on the classification model or its hyper-parameters based on cross-validation results
(Best results).
4 Association Rule Mining (4 marks)
You are required to complete this part using suitable package for association rule mining.
Code ? Analyze the dataset using association rule mining;
? Choose suitable threshold for confidence, support and/or other parameters.
Report ? Your code design and running results.
? Your findings on the association rule mining results.
Page 3 of 6
SIT742 (Modern Data Science)
Full Marks: 40
Assessment Task 02
2021 Trimester 1, Due: 8:00pm AEST, 22/05/2021
Part II
Data Analytic — Prediction (8 marks)
Google Scholar is a web service that indexes the metadata of research articles on many scientists. Majority
of computer scientists use Google scholar to track their publications and research development. Therefore,
the web crawling on Google Scholar can provide the citation information on all professors with a public
Google Scholar profile. After the crawling, the prediction could be conducted to predict the future citation
numbers such as citation all.
Task Description
In 2021, to better introduce and understand the research works on the professors, the university wants to
perform the citation prediction for individual professors. You are required to implement a web crawler to
crawl the citation information for A/Professor Gang Li from 2003 to 2021 (inclusive), and also conduct
several prediction as required. You will need to make sure that the web crawling code and prediction code
meets the requirements. You are free to use any Python package for Web crawling and prediction by finishing
the following tasks.

  1. Crawl the citation information for A/Professor Gang Li from 2003 to 2021.
  2. Train Arima on citation information from 2003 to 2017, and predict the 2018, 2019 and 2020 citation
    information. You need to draw the line plot 2
    to show the predicted citation for comparison (more
    details in below sections).
  3. Conduct the grid search on Arima parameters (p, d and q) to select the best parameter values and then
    use them to predict the citation information from 2021 to 2022. You also need to draw the prediction
    for comparison (more details in below sections).
  4. A/Professor Gang Li citation Information Extraction
    You will need to import the suitable (or your chosen) web crawling library and use the corresponding library
    to crawl the year 2003 to year 2021 citation information (19 years) for A/Professor Gang Li’s google scholar
    profile page: https://scholar.google.com/ci... Eg: citation on year 2020 is
  5. and citation on year 2021 is 228 3
    .
    5.1 Crawl and Generate the citation dataframe (1 mark)
    The code must contain the necessary web crawling steps and necessary data saving steps. The results of
    the code running will generate the citation2003-2021.csv. The citation2003-2021.csv will contain the
    year column and citations column. Data extraction without web crawling steps in the code will incur 0 mark.
  6. Train Arima to predict the 2018 to 2020 citation
    In this part, you need to train the Arima, perform the prediction and also evaluation.
    6.1 Train Arima Model (1 mark)
    You will need to use the crawled citation2003-2021.csv and then perform the Arima training with
    parameter of p = 1, q = 1 and d = 1 on data from 2003 to 2017 (15 years).
    2
    for some coding example, please check https://stackabuse.com/matplo...
    3Hint: In the right corner of Google profile page, there is a hyperlink VIEW ALL. By clicking the hyperlink, you could see all
    the citations from 2003 to 2019
    Page 4 of 6
    SIT742 (Modern Data Science)
    Full Marks: 40
    Assessment Task 02
  7. Trimester 1, Due: 8:00pm AEST, 22/05/2021
    6.2 Predicting the citation and Calculate the RMSE (1 mark)
    Then you will need to use the trained Arima model to predict the citation on year 2018, 2019 and 2020.
    You will need to perform the evaluation by comparing the predicted citation from 2018 to 2020 with the
    true citation from 2018 to 2020 and calculate the root mean square error (RMSE).
    6.3 Visualization for comparison (1 mark)
    You will also need to use matplotlib to draw the line plot with training data from 2013 to 2017, the testing
    true value, the prediction value and also the confidence interval.
    Note
    You will need to complete the notebook code, and insert the related self-written code and required results
    into the corresponding place of the report SIT742Task2-Report.pdf.
  8. Parameter selection and 2021-2022 Prediction
    In this part, you will need to conduct the grid search with Arima and select the best parameter values to
    predict the citations on 2021 and 2022.
    7.1 Grid Search (2 mark)
    You will need to run the grid search for parameters from the range p = [1, 2], q = [1, 2], d = [1, 2] with
    training data (year 2003 to 2017) and testing data (year 2018 to 2020). The result of the search on each
    parameter combination (eg: p=1, q=1, d=1) will need to be stored in the search-results.csv. The
    search-results.csv will have the column of “RMSE” and column “parameter-set”.
    7.2 Select the best parameter values and Predict for 2021 and 2022 (2 marks)
    You will need to perform the training with Arima on data from 2003 to 2020 with best parameter values
    you have found above, and then conduct the prediction for year 2021 and 2022. You will also need to use
    matplotlib to draw the line plot with training data from 2013 to 2020, the predictions 2021 to 2022 together
    with their confidence interval.
    Note
    You will need to complete the notebook and insert the related self-written code and required results into the
    corresponding place of the report SIT742Task2-Report.pdf.
    Part III
    Self Reflection - Essay (12 marks)
  9. Self Reflection Essay
    Based on your experience with the assessment tasks, you are required to write an essay with 1200-2000 words
    to reflect your understanding and thoughts on the Big data, which should include the following information:
  10. What are the Python packages that you find useful in manipulating and analyzing Big data? You can
    briefly analyze their advantages and disadvantages;
    Page 5 of 6
    SIT742 (Modern Data Science)
    Full Marks: 40
    Assessment Task 02
  11. Trimester 1, Due: 8:00pm AEST, 22/05/2021
  12. What are the Big data platforms that can help storing, retrieving and analyzing the big data? What
    are their advantages and disadvantages?
  13. Compare and contract the Python data analytical packages and their Spark packages.
  14. What are your opinions on the privacy issues in the Big data era? Any example to further illustrate
    the risks?
  15. What are the methods you think could help to solve the privacy issues on big data? Please list any
    successful implemented method.
  16. Any other thoughts about data science, or suggestions to future students (or teaching team) about this
    unit.
    Referencing should be in Harvard style, and more information about essay writing can be found at:
    ? https://www.deakin.edu.au/stu...,
    and
    ? https://www.deakin.edu.au/stu...
    ? https://www.deakin.edu.au/stu...
    introduction

    推荐阅读


    上一篇：投稿|Q1平稳度过、Q2下滑成定局，理想汽车命系新车

    下一篇：人工智能如何帮助自然科学