[SOLVED] R database graph ADM 4307 Business Forecasting Analytics

ADM 4307 Business Forecasting Analytics
Forecasting Contest
Instructions:
ADM 4307 Business Forecasting Analytics Fall 2019
Forecasting Contest
1. This contest can be done in a group of two or three students.
2. Make sure your submitted project report is neat, readable, and well-organized. Project marks will
be adjusted for sloppiness, poor grammar and spelling as well as for technical errors.
3. Front page of the report has to include the title of the project, course code and section, student names and student numbers. The second page is the statement of integrity that must be signed by
all team members.
4. Plagiarism on contest will not be accepted.
5. Questions related to the contest should be sent to the Teaching Assistant, Zahra Abtahi
([email protected]).
6. The contest report is to be submitted electronically as a single Word/PDF document file as well as
an R script file (if any) via Brightspace by Monday December 02nd prior to 18:00.
Total 100 points | Weight: 10% of the final mark | Bonus: 5% of the final mark for the winner team
Forecasting Radiation Therapy Volume for Head and Neck Cancers
A. Objective
The main goal of this project is to develop a model that can suggest ways for a provincial cancer agency to predict radiation therapy (RT) starting volumes, by month, for head and neck cancers.
B. Introduction
The provincial cancer agency deals with thousands of patients each year, providing a variety of services including conventional cancer treatments such as RT and surgery. Demand for these treatments, and specifically for RT, tends to include considerable variation across time.
High variability in demand for RT leads to inefficiencies in the system. During periods of increased demand, patient waiting times increase. This fact is an observation by the provincial cancer agency, but is also easily proved with the aid of Littles Law. Increased waiting times can cause psychological discomfort
Fall 2019 Page 1

ADM 4307 Business Forecasting Analytics Forecasting Contest
for patients and their families, and can reduce the effectiveness of treatment by allowing the cancer to progress further before beginning RT. Conversely, during periods of lower demand, resources may be under-utilized. Currently, the provincial cancer agency tends to deal with increased demand when it arises as well as they can. Being able to make reasonably accurate predictions about future RT starting volumes from existing data, however, would give them a huge advantage in planning for busy periods and in making their system more efficient.
The provincial cancer agency believes that cancer tends to occur in the population at a fairly steady rate. Yet the demand for RT that they see is quite variable. Thus, gaining an understanding about the relationships between RT starting volumes and some other variables could potentially help the provincial cancer agency work on ways to reduce this variability and ultimately improve the utilization of expensive resources while reducing overall waiting times.
In this project, we seek to help the provincial cancer agency predict future RT starting volumes particularly for head and neck cancers.
Head and Neck Cancer
Head and neck cancers correspond to cancers of the mouth and upper air and food passages (the oral cavity, oropharynx, nasopharynx, hypopharynx and larynx), the salivary glands and thyroid, and account for about 6% of all human cancers.
There are about 4,600 new cases in Canada each year.
It is more common in men than in women (the ratio is about 3 to 1).
It is more common in people over 50.
The most common sites for this type of cancer are the larynx (1/3 of cases), the oral cavity (mouth and throat) and the pharynx, and usually arise in the mucous membrane that lines the airway and food passages. The majority of head and neck cancers that arise in the mucous membrane of the mouth and throat (squamous cell cancers) are caused by tobacco smoking and/or alcohol consumption.
Traditional treatments for this type of cancer are surgery, radiation, chemotherapy or a combination of these modalities, being radiation therapy recommended as the primary method.
Radiation Therapy
Radiation therapy (RT) is a cancer treatment that uses radiation (high-energy rays) to kill or shrink tumour cells. It is used to treat some, but not all cancers. RT destroys cells either directly or by interfering with cell reproduction (normal cells are able to recover from radiation damage better than cancer cells).
Used alone, RT can be curative in many cases. It is also used in combination with other treatments/therapies such as surgery. It might be used to both reduce the size of tumours before surgery and to destroy any remaining cancer cells after surgery. When a cure is not possible, radiation therapy can also help alleviate symptoms such as pain, and improve quality of life for patients.
RT is the principal treatment for various skin cancers (melanoma, non-melanoma), cancers of the mouth, nasal cavity, pharynx and larynx, brain tumours, and many gynecological cancers, as well as lung and prostate cancer.
C. Data Available
Data taken from 8 years of history from the provincial cancer agencys patient database are available. This data set includes information on cancer diagnosis dates, admission dates, RT treatment
Fall 2019 Page 2

ADM 4307 Business Forecasting Analytics Forecasting Contest
start dates, surgery dates, information about patients, information about the type of cancer (including tumour and histology), and other potentially interesting information.
Data have been aggregated by month into 96 records (eight years with 12 months each). Each record includes the total number of diagnoses, site admits and RT starts during that month. Each record also includes the ratio of the site admits for that month that fall into various categories. For example, one of the categories is female. In January 1996, there were a total of 76 site admits, so numadmit = 76. Of these, 19 were female, so the variable called female = 0.25 (i.e., 19/76).
Table 1: Data Dictionary
#
Variable
Definition
Variables generated
1
gender
male/female
female
2
age at diagnosis
patient age when diagnosis was made
age1, age2, age3 (50, 51-70, >70)
3
diagnosis date
date at which diagnosis was made
numdiag (by month)
4
site admit date
date at which case was registered with the agency
numadmit (by month)
5
site
a code for the actual tumour site
c0, c1, c3, c4, c7 (aggregated from more detailed categories)
6
histology
histology code (related to the tissue structure)
hxxx, where xxx is the first three digits of the histology code
7
tumour sub group
code for the tumour sub group
ea, la, lp, nc, oc, ot, ph, sl, th
8
cancer centre
patient cancer centre
cc1 (Centre A), cc2 (Centre B), cc3 (Centre C), cc4 (Centre D)
9
health region
patient geographical region
hr1 (Region 1), hr2 (Region 2), hr3 (Region 3), hr4 (Region 4), hr5 (Region 5)
10
RT start date
RT start date
numrt (by month)
11
surgical date
patient surgical date,
not included
D. Some Hypotheses
The provincial cancer agency believes that RT starting volumes can be predicted by looking at recent admits, diagnoses, and various patient and cancer characteristics. The agency postulates that:
E.
1. RT starting volume (in months) can be predicted from RT starting volume from 12 months prior, 24 months prior, and 36 months prior. In other words, that the RT starting volume has a monthly seasonal component.
2. RT starting volume can be predicted from site admit numbers from recent months.
3. Site admit numbers are a better predictor for RT starting volume than diagnosis numbers.
4. RT starting volume can be predicted more accurately if, in addition to site admit information,
demographics data (age, gender, cancer centre, geographic region) are included.
5. RT starting volume can be more accurately predicted with the addition of information specific
to the cancer (tumour subgroup, histology, cancer site).
Deliverables
Please provide a report describing your approach and including a monthly forecast of RT starting volumes for January 2004 to December 2004.
Fall 2019 Page 3