Hello, if you have any need, please feel free to consult us, this is my wechat: wx91due

MAS8502: Project

Requirements

You should submit your project to Canvas no later than 12 noon on Tuesday 14th January. You should submit your report as a pdf no longer than 12 pages, along with a .Rmd file which  contains the code used to generate your analysis (i.e. submit two files).

Project brief

In this project, you will analyse the Telemarketing data set, which describes the outcomes of telemarketing attempts by a Portuguese bank to convince clients to take out long term deposits.

You should download the telemarketing.RData file from Canvas and save it to a suitable folder to contain your work.  To import the data into R, set your working directory to be the folder containing the telemarketing.RData file, and run the following line in your R session:

load("telemarketing.RData")

The variables included in this dataset are:

• age – The age of the client;

• marital – The client’s marital status, coded as 1 for married, 0 otherwise;

• default – Does the client have credit in default?  Coded as 1 (yes) & 0 (no);

• housing – Does the client have a housing loan?

• contact – How was the client contacted?  Coded as 1 for by mobile phone, 0 for landline;

• campaign – Number of times the client was contacted this campaign;

• previous – Number of times the client was contacted before this campaign

• emp.var.rate – Quarterly employment variation indicator

• You should consider a variety of different modelling approaches with different approaches to variable selection – where possible you should make sure to explain the model design decisions  you  have  made.   Again  this  is  a  real  dataset,  and  so  there  will  be  several perfectly valid approaches here, just be sure to explain what you have done and why you have done it.  You ought to consider at least one method of regularised regression and compare and contrast regression method(s) with those from discriminant analysis.

•  Make sure to include a discussion of inferences from your models as well as how well they predict – which variables were/weren’t useful predictors of whether the campaign would be successful? What was the effect of the variables which were significant?

•  Consider how you can report and compare the effectiveness of your different classifiers in a way that is fair. Which model performs best? What are your conclusions?

Writing the report

• You should experiment with different plots and commands using the console directly (or in a .R file that you use for exploration).
• However,  when you’re creating the report,  you must put the text that describes you findings and the R code required to produce the plots or numbers in a single R Markdown file (with extension  .Rmd)
• The main exceptions are the commands setwd(), View() and    (potentially) install.packages(). Never  include  these   commands  in  an  R  Markdown  file  as these are details that are specific to each person’s computer.
• Do not pre-generate the results, save the plots, and then manually put the plots into a Word document (or equivalent) that contains the text.  The plots and numbers should be generated automatically when you click the Knit button in RStudio.  This is a much more efficient approach.
• Related to the above, do not generate a Word or HTML document using the .Rmd file and then print it as a PDF. Again, this is the wrong approach because it is less efficient and because the formatting of HTML and PDF files is deliberately quite different (you wouldn’t expect a web page to look the same as a printed report).
• One option that is useful if you want to hide code  (e.g., due to a page limit) is to set echo = FALSE in  knitr::opts_chunk$set(echo = FALSE).  For  this  assessment,  you don’t need to show your R code in the PDF because you have to submit the Rmd file as well, where I will be able to see your code.

Marking criteria

Reports will be marked on the university scale.  Credit will be given for:

• Mathematical accuracy – How well you apply the statistical techniques in your report.
• Methodology – An understanding of why you have chosen the techniques that you have, and what their output means in terms of your investigation.
• Critical evaluation – A discussion of the strengths and weaknesses of your methods, how things could be improved, etc.
• Report structure and presentation – How well your report is written in terms of structure, how well it flows, and so on (i.e., aiming for a single, coherent piece of writing, as opposed to lots of separate answers jammed together)
• Reproducibility – Can I reproduce your report using the .Rmd file?  You can check the repro- ducibility by copying your files to another computer, and see if RStudio generates the same report without errors.
• Extra  credit will be given for any reading/techniques  implemented  from  outside  the scope of the module, but this is not a requirement to receive a good mark.  Similarly, any investigations carried out beyond what was described in the report task will also be considered for extra credit.

As such, the marking scale (out of 100) will be:

•  80+ (Upper distinction) – A publication quality piece of work.

•  70 − 79 (Lower Distinction) – A very good piece of work showing strong understanding.

•  60 − 69 (Merit) – A solid attempt, mostly hitting the main points.

•  50 − 59 (Pass) – More good than bad, but clear areas to improve.

• < 50 (Fail) – Significant things went wrong or weren’t attempted.