Home Computer Science
Structuring Successful Internships
As a high-impact practice, internships are a common requirement in many undergraduate programs besides data science, and there are common aspects to all successful internships. Any internship program should prepare the student for success after the internship is completed. The training that undergraduate students in data science undertake in the first few years of their program is broadly dispersed between mathematics, statistics, and computer science. This means that when it comes to structuring internships for data science students, there are many potential types of projects where a student can make a meaningful contribution. For example, in our program, several students have had successful internships in what are probably best described as “data-centric software engineering” roles, focused on tasks such as building databases to integrate and clean up company data. Other students have focused on predictive modeling. As an example, one student from our program recently completed a successful internship developing predictive models for insurance data using neural networks. A key contributor to the success of an internship is ensuring the suitability of the chosen role for the students: the student working in the “data engineering” role was one of our strongest programmers, with excellent software development skills, while the student working in the “data science” role building neural networks came in with a deeper than average mathematics background that prepared him to take a role building sophisticated models before he had been thoroughly trained in them through our curriculum. It perhaps goes without saying that managers should seek interns with the right skills for the job; although all undergraduates in analytics and data science go through the same coursework and develop a similar skillset, some will be more suited for some roles than others.
Keep in mind that from the student’s perspective, accomplishing the technical aspects of the project are only a part of the overall objective. The best student internship experiences will provide opportunities for the student to network and establish contacts in the field, as well as to present their work and build a portfolio that they can share as evidence of work accomplished. Such opportunities are especially important in the field of data science and should be a part of any internship experience.
Structuring Successful Capstones
The capstone project is often the final requirement in an undergraduate degree program. The biggest difference between the two is that capstones are commonly offered as “courses” with 10-20 students working together, while internships are commonly offered to an individual student who will work independently under the supervision of a faculty member. As with internships, there are common best practices for capstone courses that analytics managers seeking university relationships should consider. For the student, the capstone project course should provide an opportunity to demonstrate both the breadth and the depth of their newly acquired skillset. Because the capstone is overseen by faculty and work is primarily done on campus, this requires the manager to be able to essentially hand off the project for a period of time (10-15 weeks for a standard 15-week semester) while the students work under the guidance of their faculty mentor. There needs to be clear communication between the manager, the faculty overseeing the capstone, and the students. Tlie manager must have a certain degree of confidence in the faculty mentor who will oversee the project. Given the degree of separation between the manager and the students, the project needs to be clearly delineated. What are the objectives? What are the deliverables? What is the timetable to completion? What data is needed? Who will provide the data, and how will it be managed? All of these questions and more need to be carefully thought through when planning a capstone.
High Impact in Action – Two Case Studies
Jeremiah shared an internship case study from the B.S. in Data Science Program at the University of New Hampshire. He provides context from the perspectives of the student, the university and of the sponsoring organization.
In the Spring of 2020, a third-year data science student interned at a local startup company specializing in in autonomous systems, unmanned aerial vehicles (UAVs), drones, and counter-unmanned aircraft systems (CUAS). ’[he company’s products included an intelligence analysis software based on telemetry data that determines a UAV’s payload deployment, threat assessments, anomaly detection, FAA compliance, mission profiling, geo-fence violation, and rollover detection. The student’s role was to develop several classification models related to product performance over a 15-week period (one semester). An example of the student’s project included developing a classification model to determine if an unmanned system travels in a straight line for the majority of the flight path. For this project, the student had to use the Pandas, NumPy, and Matplotlib libraries within Jupyter Notebook to analyze the latitude and longitude values within an unmanned system flight path. The student also had to work with other unique data, such as UTC timestamps and battery voltages while using GitHub for code revision over several development iterations.
During the three months of the onsite internship, the student attended daily staff meetings, biweekly check-in meetings with the director of engineering, and monthly staff meetings, where the CEO and COO would update the staff and contractors regarding company accolades, developments in their market, and future plans.
The internship was successful and mutually beneficial for both student and company. The aspects of the internship that contributed to success include:
■ The objectives of the student project were clearly defined, with a detailed calendar of weekly expectations and deliverables.
■ While the student had limited domain knowledge (few undergraduates will have any industry knowledge when they begin an internship), he was well versed in the data science-dimensions of the project and was highly proficient in the programming languages required.
■ The student had a direct supervisor for the duration of the project who provided feedback and mentoring (and infinite patience).
■ Through staff meetings, the student was able to engage with other employees, senior leaders, and other parts of the organization, which allowed him to better understand the organizational culture.
■ The project was substantive and relevant to the business - it was not “busywork”.
The second undergraduate case study highlights an applied project course. This case study comes from Dr. Riaan De Jongh, Director of the Centre for Business Mathematics and Informatics at North-West University in South Africa.
When students arrive at the university for their first year of study, they have a vague idea about the problems or tasks they will encounter in their professional careers. Frequently students enquire about the type of problems they will face one day, and how the course/modules they are taking will assist them in solving these problems. To give them a flavor of what they could encounter in practice, a first- year course was designed with the main objective to expose the student to a purposefully vague financial problem statement (provided by a large regional bank), which they have to formulate and then solve using a combination of mathematics and computer programming. In the process, the students should experience initial bewilderment and the pain of not knowing where to start. They should realize upfront that they do not yet have the necessary knowledge and are not able to ask the right questions to formulate the problem. Then the students are taken through a process of acquiring the relevant knowledge which will eventually enable them to ask the right questions to formulate the problem in mathematical terms. Once this is achieved the student is required to derive the mathematical solution to the problem and translate it into an algorithm that can be programmed. Part of the process involves confronting the student with other questions relevant to the problem at hand. In this way the student explores other facets of the problem that increase the functionality of the decision support system they have to design, develop, and present to the sponsoring financial institution. Although the students have the benefit of the lecturer’s guidance on the financial concepts, the various problems to be addressed, the associated mathematical formulations and resulting equations, very little guidance is given on the design of the decision support system and the associated programming requirements. The case study addresses all the ingredients of a typical data science project, albeit without the typical large data set, which they will experience in later, more advanced courses. The course requires students to integrate mathematical problem formulation, computer programming, creative thinking, project management, system design and development, and presentation skills. A “typical” course problem:
1. Initial problem statement
Your fellow student would like to prepare for retirement, and you need to advise him how he should plan for this. In order to do proper financial consulting, you have to design and build a decision support system. Questions that your fellow student could ask includes, but are not limited to: How much money should I set aside now and until my retirement, on a monthly basis, to ensure that I will make adequate provision for my retirement?
2. Detailed problem formulation
After your interview with the student, you have the following information:
With the information provided, calculate the value of the first payment at the end of January 2021. Develop an excel sheet where you can illustrate the effect on the first payment when there are changes in the input parameters. Draw graphs to illustrate this. What parameters have the greatest influence on the initial payment? Will you be able to handle once-off payments and/or withdrawals by the student? Think about the functionality you would like to build in your decision support system.
3- Mathematical formulation
Formulate the requirements mathematically. Test the formulas and report the results.
4. Decision support system
Your decision support system should at least be able to solve the mathematical formulation but should be flexible enough to answer most of the problems that you had to solve in this module. Although you have done most of the practical work in MS Excel, you may use any programming language to develop your system; for example, Python, Visual Basic, C##, R, SAS or any other programming language you are comfortable with. You may even consider developing an application in Apple IOS or Android. The most important thing is that you use your creative thinking ability to come up with a system that is able to address all sorts of other questions. Consider some of the questions we asked you in the exams and assignments. Would your system be able to solve these? Examples of typical questions are given below.
i. For three months the monthly effective interest rate is 0.90% per month. It is followed by a nominal interest rate of 8.8% per year, compounded quarterly for the following nine months. Calculate the equivalent continuous interest rate over the period.
ii. For six months the interest rate is 8.5% per year. It is followed by an interest rate of 4.6% effective per half-year, for the following year and a half. Calculate the equivalent nominal interest rate, compounded monthly, over the period.
On 1 July 2018, Gianni opens a savings account and starts to invest an amount of money at the end of each month into the savings account. The first investment is $150 and will increase every month thereafter with 0.7%. He plans on making these investments for a total of seven years. On 1 January 2021, Gianni decides that for the next three and a half years he will make an additional constant investment of X at the start of every quarter into the savings account. The savings account earns interest at 6.8% per year, compounded quarterly. Calculate the total present value of all the investments in the bank account on 1 January 2017-
A young couple decides to buy a house for $100,000. They plan on paying 15% of the purchase price immediately as a deposit. The remaining balance of the purchase price will be financed by a bank loan that will be repaid over a period of 20 years. They agree to repay the bank as follows:
The interest rate is constant at 3-5% per year.
i. Calculate Y.
ii. Calculate the interest and capital component of the 121st payment.
5. Final deliverable
Students must develop a deployable decision support system. An example of the front end of such a student’s final deliverable is presented in Figure 3-3- The decision support system is characterized by many functions including, but not limited to:
The system depicted in Figure 3-3 supports the ability to solve all three questions.
This applied project course at North-West University is regularly over-enrolled and is one of the more popular courses on campus, with many students ultimately working for the banking institutions that provide problems for the course. Students eventually take a more advanced capstone course at the end of their program, which
Figure 3.3 Example of undergraduate data science students' decision support system.
builds on the “high impact” practices from this first-year project course. The specific aspects of this course that contribute to its ongoing success include:
■ The professor has a long, successful track record of teaching the course and helping students with placement after graduation. As a result, he, and other lecturers that present the course from time to time, have a strong reputation and credibility within the local business community. They understand the needs and “language” of the business community and can translate this into an academic curriculum.
■ While the “content” of the problem may change every semester, the course project process does not change.
■ Given that this course is housed within the Business Mathematics and Informatics Centre at the University, it would be expected that the students are computationally strong. This project course emphasizes “softer skills” such as problem formulation, working in teams, communications, and presentation skills, which may not come as easily for the students.