Ref1
docx
keyboard_arrow_up
School
University of South Florida *
*We aren’t endorsed by this school
Course
BADM533
Subject
Industrial Engineering
Date
Jan 9, 2024
Type
docx
Pages
4
Uploaded by HighnessCaribou1867
Reflection 1
The goal of this reflection is for each student (individually) to assess the quality of their team’s
Homework deliverable. The response to each section can be accomplished in
1-2 paragraphs
.
Before writing a response for each section,
review the Homework rubric, and be sure to evaluate
your team’s submission along each of the deliverables’ requirements.
Your objective here is not
to defend the submission, but as an impartial observer, critique and assess its quality. Therefore, your
reflection should have a narrative (persuasively supported by evidence in the submission). Grading of
this reflection is based on how well you demonstrate the following:
Critical thinking
- shows original thought, logical organization, cohesiveness, and both
accurate and critical evaluation of the deliverable(s)
Specificity
- statements of evaluation/quality of the deliverable(s) include references to (and
the locations of) specific artifacts, results, analyses, or conclusions that provide evidence in
support of the specific evaluation
Clean execution-
free of spelling error and typos, does not appear sloppy or thoughtless, all
elements of the presented work are polished
Direct Questions
(Please answer honestly. The direct questions, i.e., the next four question will not affect your grade, but
help us assess the class, and may help you
with your reflection below.)
Do you think that the homework reflected the topic and skills covered in class?
Yes
Did the homework help you improve your understanding and skills, in reference to the topic and skills
covered in class?
Yes
Approximately, how much did you, as an individual, spend on the homework?
5 hours
Approximately, how much did your team spend working together on the homework?
15 hours
Assess your team’s problem-solving approach
The teams first step to approach the problem was to understand the context of the problem and the
application/s of the solution that we will come up. We individually came up with information around
the problem context and what we think might be the expectations of the client and how will the client
measure the success of the recommendations once applied. We tried to answer questions such as
“What are the factors that the Coach and players work on regular basis? What all factors helps a
team win, lose or tie? How is the progression of team decided in the soccer league? What all data
points a coach might be interested in when s/he decides the strategy of the team?”. We did not think
through the application journey of the recommendation at first and only after a few discussions took
the approach.
As team we did not evaluate the implications of the problem context and the solution before starting
our analysis. We did not consider how our recommendations might go wrong or our analysis might
give no tangible results. This might have caused the entire analysis to yield no result. Going forward
we will have to do a sample analysis to find out whether the recommendation we are aiming for or the
analysis we are aiming for will generate a tangible result. This not just helps us put a structure to the
problem but also creates sense of the direction in the approach.
Post this step we moved on to analyze the data given to us and how each of the factors given in the
data can be used to build a recommendation. After analyzing the data, we combined it with the first
step to frame and narrow the problem statement, “How can we decrease the loss percentage of AS
Roma in the Italian league?”. This was an iterative process because we would start with the problem
statement, evaluate the information in the data in context of the problem statement and evaluate the
application and implication of the possible recommendations. Problem statement such as “ How AS
Roma can improve their performance by changing the players attributes or by inducting players with
certain key attributes?” seem good but the application was limited. Even the problem statement such
“How AS Roma can win more matches?” was very abstract when it came to the application aspect.
Working iteratively we zeroed in on the problem statement and further narrowed it down to “How can
we decrease the loss percentage of AS Roma in the Italian league by recommending a winning
strategy for matches where AS Roma lost by a margin of 1 goal or tied after scoring goals ?”.
Here we sampled our data to restrict our analysis to Italian league which might have given a biased
data as the player attribute or the strategy against the players is valid across the teams and leagues.
Usually players play in multiple leagues and their attributes are same for different leagues. Having a
larger data set enhances the performance of the model. The downside of which is that some of the
leagues are drastically different from others and players might change their strategy. We did not
analyze this aspect of the data.
In step 3 we created the required data for our analysis by filtering out the matches in which AS Roma
lost and what strategy they followed. We presented this analysis in the data exploration section and
backed our analysis by citing the number of matches AS Roma lost and how a win even 10% of those
matches can help achieve better performance ranking for AS Roma.
We selected the top 5 teams (as they accounted for more than 60% of the loss or tie of AS Roma)
against which AS Roma had either lost or tied with and looked up the teams against which these 5
teams lost frequently. This gave us a data set to extract a winning strategy against these teams and
how these strategies were different from the strategy of AS Roma. To make the recommendation
actionable we looked at the strategy of the teams which looked like AS Roma in terms of the team
composition.
We collected the strategies and then found out the association between these strategies and the
performance delivered at a confidence of 0.01 and confidence of 0.5. The association rule mining
gave us close to 2500 rules and we plotted the rules to shortlist the rule that delivered the most
number or performances.
We looked up on the web to find out what does the strategy mean and how do teams interpret it and
why that strategy might help AS Roma. We found that the strategy was a defensive one which meant
that the teams should focus on not conceding the goals by deploying more players in the defense.
This is not a popular strategy but works for the weaker teams when they are playing against very
strong teams. This helps them reduce their losses but might not be helpful in helping them win.
Assess your team’s technical deliverable –
In technical deliverables we included interpretation of each of the steps reported and how the
outcome and the execution of that step is relevant to the problem and how it logically adds up to the
solution approach but the flow of output from one step into the input of the next step and it’s
interpretation was missing from our report. We could have included that to make the report and
analysis more clear.
In technical deliverable we presented the Data Exploration, Data Cleaning, Feature Engineering and
Model building as part of our analysis and report.
Data Exploration -In data exploration we included the steps we used to get the structure of data, the
distribution of data and the anomalies in the data. To achieve this, we individually explored the data to
find out the data structure, the patterns, the possible anomalies and interpretation of the data in
context of the problem. We checked the Kaggle kernel to see different ways in which people have
explored he data and are there any specific libraries and methods that we might consider for our
analysis. In the report we have included our interpretation of the data explored and the process of the
exploration.
We could have avoided the excessive results and outputs which are not relevant or give any tangible
information to the reader. This increased the length of the report and was of no value to the end user.
Data cleaning- Once the data was explored, we cleaned the data for the analysis. The data given to
us was clean and well-structured so not much effort was required at this step. We looked at the data,
both categorical and continuous variables to check if the distribution was logical and mapped to the
real-world interpretation.
Here we interpolated the data where the rows were missing using forward interpolation, but we could
as well have avoided the steps and removed the step as the proportion of missing data was very less.
Feature Engineering – This was an important step in our analysis as the data given had close to 150
attributes. We filtered for the target variable and the attributes we were looking at for our analysis and
to build our recommendation. The step required some data transformation such as transforming the
players position attribute into coordinate system and combine it with the other players to build the
teams playing strategy. We combined the player attributes to build the team attribute and that gave us
the data points to build the model.
While doing the feature engineering we took the cut off values based on the median and that could
have been done better by using the actual values by looking at the AS Roma’s actual cutoff values for
deciding their formation. This would have made the analysis more real.
At each of the steps, we included the following information to ensure that there is logical flow of the
report and interpretation:
1.
Why we were doing that step – here we included the information and our reasoning for
why we were including that step, this included the relevance of the step in the business
context, the information we were looking for.
2.
The result of the step – what does the result mean, how it explains our next step and how
it fits our initial hypothesis.
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
- Access to all documents
- Unlimited textbook solutions
- 24/7 expert homework help
To enhance the interpretation and readability of the technical report we could have talked about how
different parameters played the role in the execution of the step and how changing that parameter
might change the overall output and the solution.
Model building – Before model building, we analyzed different hyperparameters of the algorithm used
for the model building and the interpretation of these hyper parameters and the resulting parameters
in context of the problem.
We stuck to the Apriori and did not try other algorithms to find out our result or cross validate our
results.
Assess how you (and the team) can improve subsequent assignments
Individual Improvement – For subsequent assignment I can improve by following methods:
Not focus on solution without creating a very detailed problem statement and working
strategy.
I would like to work on the assignment daily for 30 mins to 1 hr to create a consistent
thought process and approach rather than spend 2-3 hours in a go.
Read through the theoretical literature of the algorithm before going for the technical
tools.
Teams improvement - For subsequent assignment I believe the team can improve through the
following strategy:
By knowledge sharing and better organization of resources much ahead of the team
meeting rather than in the meeting so that every team member gets an opportunity
to read it and come to the meeting with an interpretation.
Start compiling the report earlier rather than wait for the completion of technical
analysis.
Provide Deliverable(s) Grade
Overall score: 28 / 30
Problem Solving Approach: 9 / 10
Technical Deliverable: 19/ 20