计算机专业essay/CSproject/report代写
Chapter 2
Case Description
2.1 Introduction
DHL LLP is a group within the DHL Global Forwarding division belonging to the Deutsche Post DHL. LLP stands for Lead Logistics Partner and acts as a neutral partner. DHL LLP instigates and manages change across the customers entire supply chain, to meet changing business and customer demands. LLP does this by bringing continuous improvement and cost reduction, introducing lean logistics processes and optimizing logistics networks and transportation flows. More will be explained about DHL LLP during the guest lecture. For this case you will consider shipment data of one of one of the biggest airplane manufacturers in the world, for which DHL LLP manages the transportation. You will be provided with a data set containing data on origins, destinations, volumes, weights, etc. for these shipments. It will be your task to use and analyze that data in order to build time series, test for seasonality, make forecasts, and analyze and improve the given clustering. You should achieve this using the techniques and the software that you know from your Econometrics and Operations Research courses.
2.2 The data set
The fileTRsFullDataUM.xlsxcontains data of 81520 shipments that were carried out during the time frame from January to October 2017. Below you find a table with the explanation of the dierent columns. You can save this file as a CSV-file in order to read the data into a Java program.
Column Column Title Content
A TR Code Unique code of transportation request
B TR Creation Date/Time Creation date of transportation request
C TR Source Location Code Unique code of source location
D TR Dest Location Code Unique code of destination location
E TR Gross Weight (KG) Gross weight (in kg)
F Nb of Ship Units Number of shipped units
G Origin Country Country of origin
H TR Source Location Postal Code Postal code of origin
I Origin City City of origin
J OriginFull Full address of origin
K OriginCluster Cluster ID of origin
L OriginClusterLat Latitude of origin cluster
M OriginClusterLong Longitude of origin cluster
N OriginLat Latitude of origin location
O OriginLong Longitude of origin location
P Dest Country Country of destination
Q TR Dest Location Postal Code Postal code of destination
R Dest City City of destination
S DestinationFull Full address of destination
T DestinationCluster Cluster ID of destination
U DestinationClusterLat Latitude of destination cluster
V DestinationClusterLong Longitude of destination cluste
W DestLat Latitude of destination location
X DestLong Longitude of destination location
Y TR Pickup - Event Day Weekday of actual pick up date
(0=Sunday, 1=Monday, etc.)
Z TR Gross Volume (M3) Gross volume of transportation request
AA PUDate actual pick up date
AB Distance Distance between origin and destination (in km)
AC ClusterDistance Distance between respective clusters (in km)
2.3 Assignments
There are in total six questions that you have to work on, the first half focuses on Econometrics techniques, the second half on Operations Research techniques.
1.Data (pre)processing [10 points]
An important task in applied econometrics is the collection of data. Sometimes it s easy. You
click and download data from OECD data base for instance. Sometime its more messy due to
the use of dierent sources, to breaks in the definition of series, to dierent frequencies. The
data set proposed by DHL has already been preprocessed. However, a bit of manipulation
should be done to obtain the time series we want you to work with.
- You first task is to find the most frequent origin cluster. For that cluster, find the two most frequent destination clusters (column T). Let s call the link between the origin and destination clusters a lane, such as you have to develop models for lane 1 and lane
- For each lane we are interested in 3 series (You consequently have 6 series to investigate):
- Column E: Gross Weight (KG),Wt
- Column F: Nb of Ship Units,Ut
- Column Z: Gross Volume (M3),Vt.
- Build time series for those lanes/series. You have daily data for six months, sometimes with more than one shipment per day. Aggregate them to have only one observation per day. Choose probably a 5 days a week frequency.
2.Daily seasonality [10 points]
DHL has observed that shipments are dependent of a particular day in a week. There might
consequently exist a seasonal pattern in the series that is worth exploiting.
Then,
- Determine/test whether you prefer to take the variables in levels, in log-levels or in growth rates. Maybe just by looking at series and not necessarily using formal (e.g. unit roots) tests. Indeed the span of data (six months) is likely too short to really trust those tests.
- Carry out regressions for the six series to determine whether there is a significant daily eect. For instance for the gross weight for lane 1, the regression will be
WtL^1 = 0 + 1 Tuesd+ 2 Wed+ 3 Thur+ 4 Frid+"t
- Test the null of no seasonality, i.e.
H 0 : 1 = 2 = 3 = 4 =
- Look at residuals and detect (test, plot ACF) whether you have some autocorrelation. Add lags of the dependent variable if needed or identify the ARMA model or add exogenous variables, etc.
- Provide final specifications with robust standard errors if you have heteroskedasticity (test for that).
3.Forecasting [10 points]
Once you are happy with your specifications, forecast the next week shipments, i.e. the week
after the last calendar point available in the data set.
Information for questions 4
For questions 46 you should implement classes and algorithms in Java that can be used
to answer the given questions. Carefully think about which and how many classes you are
going to use before you start implementing. This part considers the assignment of origin and
destination locations to the respective origin and destination cluster. In order to reduce the
number of shipments for testing your algorithms, pick up dates will be neglected. Therefore
you should, in a pre-processing step, aggregate the shipment data (weight, volume, number
of shipments) for a given origin-destination pair over the given time period. This can be done
either using Excel or using the Java data structures that you will develop for question 46.
4.Clustering I [10 points]
Develop an appropriate data structure that will allow you to analyze the clustering of the
shipments that is given in the data set. Your data structure should be flexible enough so
that it can be extended to change the origin or destination cluster assigned to a shipment or
similar other changes. Read the remaining questions before finalizing your data structure.
Using your data structure, implement methods that can compute the following performance
measures of the current clustering:
- the total distance of all shipments from the origin location to the origin cluster,
- the total weight that is being transported from the origin locations to the origin clusters,
- the total volume that is being transported from the origin locations to the origin clusters,
- the total number of shipments that is being transported from the origin locations to the origin clusters,
- the same four measures as above, but for the destination clusters.
Summarize the above measures for the current clustering by listing the values for these
measures for the average cluster as well as the three clusters having the lowest (highest) such
values.
5.Clustering II [10 points]
Develop and implement an algorithm that can find a dierent assignment of the locations to
clusters, so that the total distance of all shipments from the location to the assigned cluster
(origin and destination) is small (not necessarily minimum). The algorithm should take as
input the numberkof clusters that can be used.
Run your algorithm for dierent values ofkand evaluate the measures from above for your
obtained solutions.
6.Clustering III [10 points]
Suggest alternative ways of clustering the shipments and implement solutions to find such
clusterings. This could for example be based on the direction of the flow, destination country,
the length of the shipments, etc. Using your approach you should be able to create distinct
groups of locations that follow a similar shipping profile.
Argue why you chose for your proposed alternative and discuss the outcomes.
7.Business Summary [10 points]
Write a two page business summary that gives a concise summary of your results. This
should be an easy to understand text for the nonexpert manager from which managerial
conclusions can be drawn and should include some figures and/or charts representing the
main results.