BST832 Forecasting
Academic Year 2019-2020 (Spring Semester)
COURSEWORK
This assessment takes the form of an essay and accounts for 100% of your total mark and the
deadline for submission is on Wednesday 20 May 2020 by 11:00am.
The Task
The objective of this coursework is to go through all steps of a forecasting task.
Once you define a problem, determine what to forecast and gather relevant data (Part A), you
are asked to check prepare your data for analysis and visualise them to determine whether they
contain any key feature(Part B), then you need to identify suitable forecasting (Part C), specify
them, estimate their parameters and check if they capture key features (Part D), finally you
need to check the validity of your model and evaluate its point and interval forecast
accuracy(Part E).
I will provide you with a template in RMarkdown that you use to write your essay.
You finalise your Rmarkdown file and then convert it into a Pdf file. You need to submit both
RMarkdown and Pdf versions.
Part A: Problem definition and gather information
Forecasting is not an end by itself. It is an input to other functions in organisations. It is
generally used to provide evidences and inform decisions regarding the future. The first step in
any forecasting task is to determine what are decisions that forecast is going to support, so you
need to start by identifying those decisions.
In Part A of your coursework, you are asked to elaborate on a problem you are interested in. It
could be in any sector such as supply chain, logistics, shipping, health, humanitarian, energy,
third sector, etc.
You need to discuss one decision or planning task that required forecasts. Following that, you
are asked to determine what you need to forecast. That involves identifying
predictor/forecast/response variable, forecast horizon, forecast unit of time, hierarchy or group
items, and anything else that is required for your decision making and planning process.
Once you know what to forecast, you can gather relevant data that enables you to provide future
estimation of the forecast. So, you need to find one or more datasets and describe them in this
part. The dataset must be publicly available, few resources to find dataset are:
1. https://www.kaggle.com/datasets
2. https://digital.nhs.uk/data-and-information/publications/statistical/hospital-episode-
statistics-for-admitted-patient-care-outpatient-and-accident-and-emergency-data
3. https://data.humdata.org/dataset
4. https://statswales.gov.wales/Catalogue
5. https://pkg.yangzhuoranyang.com/tsdl/
Part B: prepare data and visualise
Once you have the data, you can start by importing them into R. you need to make sure your
data is a quality data and if required you need to clean data, remove duplication, deal with
explicit and implicit missing values.
You need to create suitable graphs to identify key features of your time series. You need to
describe and interpret key features available in your data.
Part C: Select a suitable toolbox of forecasting models
Your toolbox should contain:
• Two exponential smoothing models, one automatic and one determined by you based
on data features
• Two arima models, one automatic and one determined by you based on data features
• One regression model.
• One benchmark model.
The selected models should be able to capture collectively different underlying time series
characteristics (level, trend, seasonality, autocorrelation). A full justification of the selected
models should be provided.
Part D: Fitting models
In this part, you need to specify models discussed in Part C and estimate their parameters. That
means you will create a fitted line per model. What does these fitted lines tell you about
models? Critically evaluate residuals of models and discuss whether the selected models are
reliable and capable of capturing key features or not.
What do you conclude in this section? Are you satisfied about the forecast accuracy of these
models? What do you propose?
Part E: Model accuracy evaluation
You need to select a strategy to decide which forecasting model you would use! Describe your
strategy and justify why it is appropriate? Evaluate the forecasts produced using at least three
appropriate point error measures and one interval accuracy measure. Evaluation should be
across horizons. Justify the selection of these error measures over other possible candidates.
Compare the accuracy of the proposed selection strategy. Which model do you select as the
most accurate model to suggest it to the decision maker for use in the future? Produce forecasts
based on the required forecast horizon and applied to the whole data set. Finally, you need to
visualise forecasts!
Please note that in all the above parts the quality of presentation, critical discussion and
appropriate references to the literature will be taken explicitly into account towards the
mark to be allocated. You need to use R software to do all analysis, any other tools such
as Excel is not allowed.
------------------------------------------------------------------------------------------------------------
The essay should be NO MORE THAN 3,000 WORDS IN LENGTH and all sources should
be acknowledged in the appropriate place in the text. You are advised to use the Harvard
referencing system. References are excluded from world count.
Submission is on Wednesday 20 May 2020 by 11:00am.
Essays must be submitted online on Learning Central BEFORE 11:00 a.m.
Note: you will be provided with a template in RMarkdown to write your essay.
References
Ensure all sources of information are referenced correctly using the Cardiff Harvard Style of
Referencing – if unsure see the handout from the library.
Unfair Practice
This is an individual assignment and you are advised not to engage in any activity that might
lead to suspicions of Unfair Practice. Details of the University Regulations may be found at
https://intranet.cardiff.ac.uk/students/your-study/exams-and-assessment/sitting-
yourexam/cheating-and-unfair-practice and you should familiarise yourself with these
regulations before starting your coursework.
Students are advised to keep a second copy for themselves. Should there be special
circumstances that mean you are unable to meet the submission deadline, you must obtain an
extension from the Chair of the Board of Examiners. Forms are available from room A-04 or
Learning Central. If you are not in Cardiff then contact your Personal Tutor.
Dr. Bahman Rostami-Tabar
Coursework marking-criteria
For 90%+
An outstanding piece of work, showing mastery of the subject matter, with a highly developed ability to analyse,
synthesise and apply knowledge and concepts. All objectives of the assignment are covered and the work is free
of error with very high level of technical competence. There is evidence of critical reflection; and the work
demonstrates originality of thought, and the ability to tackle questions and issues not previously
encountered. Ideas are expressed with fluency. All coursework requirements are met and exceeded.
For 70% - 89%
An excellent piece of work, showing a high degree of mastery of the subject matter, with a well-developed ability
to analyse, synthesise and apply knowledge and concepts. All major objectives of the set work are covered, and
work is free of all but very minor errors, with a high level of technical competence. There is evidence of critical
reflection, and of ability to tackle questions and issues not previously encountered. Ideas are expressed clearly.
However the originality required for a 90+ mark is absent. All coursework requirements are met and some are
exceeded.
For 60%-69%
A very good piece of work, showing a sound and thorough grasp of the subject-matter, though lacking the breadth
and depth required for a first class mark. A good attempt at analysis, synthesis and application of knowledge and
concepts, but more limited in scope than that required for a mark of 70+. Most objectives of the work set are
covered. Work is generally technically competent, but there may be a few gaps leading to some errors. Some
evidence of critical reflection, and the ability to make a reasonable attempt at tackling questions and issues not
previously encountered. Ideas are generally expressed with clarity, with some minor exceptions. All coursework
requirements are addressed adequately.
For 50%-59%
A fair piece of work, showing grasp of major elements of the subject-matter but possibly with some gaps or areas
of confusion. Only the basic requirements of the work are covered. The attempt at analysis, synthesis and
application of knowledge and concepts is superficial, with a heavy reliance on course materials. Work may
contain some errors, and technical competence is at a routine level only. Ability to tackle questions and issues not
previously encountered is limited. Little critical reflection. Some confusion and immaturity in expression of
ideas. Most coursework requirements are addressed.
For 40%-49%
A poor piece of work, showing some familiarity with the subject matter, but with major gaps and serious
misconceptions. Only some of the basic requirements of the work set are achieved. Little or no attempt at analysis,
synthesis or application of knowledge, and a low level of technical competence, with many errors. Difficulty in
beginning to address questions and issues not previously encountered. Some intended learning outcomes are
achieved.
For 30%-39%
Work not of passable standard, with serious gaps in knowledge of the subject matter, and many areas of confusion.
Few or none of the basic requirements of the work set are achieved, and there is an inability to apply knowledge.
Technical competence is poor, with many serious errors. The student finds it difficult to begin to address questions
and issues not previously encountered. The level of expression and structure is very inadequate. Few intended
learning outcomes are achieved.
Below 30%
A very poor piece of work, showing that the student has failed to engage seriously with any of the subject matter
involved, and/or demonstrates total confusion over the requirements of the work set. Virtually none of the
intended learning outcomes are achieved.