Problem Solving as a Data Scientist

Critically think about your role as a data scientist and what you have learned in this program as you answer the following prompts: A current business problem that your organization faces that you can work towards solving with the skillsets and experiences you’ve gained in this program? If you are not currently employed, what is a different real-world problem that you might solve? What technologies / languages need to be “brought to bear” on the problem?

One of the problems that I can see my organization facing in the future is product testing, metric measuring, and forecasting. I work in a project office in army aviation. At some point we test new features and new air vehicle models. I have already discussed working with the test engineering team to build dashboard to display test metrics and help with predictive models. This program has given me the foundation needed to assist with those topics. It has also allowed me to be more deeply invested in my job and use my skills in more areas than one. I am also currently a data manager so I put together a lot of reports and plans for how we handle all forms of data in our office. This program has already helped me to run reports more efficiently and provide better insight into issues with data.

I work as a ski lift operator at Breckenridge Ski Resort, under the umbrella of Vail Resorts, in Breckenridge, Colorado.

An interesting business analysis that I could work with using tools in this program is something I have actually spoken to the senior manager of mountain operations about. Our mountain has an app that gives updates on all facets of the resort. It provides live updates on the wait time of each lift on the mountain, volume of people inside the restaurants scattered across the mountains, snow conditions, and various other points of interest. The way in which wait time is calculated and predicted is very interesting. Using data from the number of passes scanned at base lifts, people actively tracking their ride on our app, and purchases at restaurants, predictive analysis can be used to provide information on the number of people out and how long it will take to get food or to ride any of the 30 lifts across the mountain. Once the information is collected, we turn to the dashboard interface on the app and can portray the data in any number of ways.

As the mountain expands and the data collection gets smoother, the predictive methods are continuing to be better and better and provide the data in a clearer manner. It was cool to discuss this, as it is not at all part of my day to day work, but something I have interest in and maybe something to work towards within the company in the future.

I work for a medical logistics company, and one area that falls under my responsibility is maintaining our handset and scanner inventory. One of the critical problems I encountered was holding a stable stock of devices for drivers in the field. There is an extremely high turnover of devices due to being destroyed, lost or stolen, or added work. Recently, I wrote some queries in SQL. These included ten years of data, geographical results of where tablets were utilized nationwide, and the growth of the business. Our company does not use Power BI, so I had to create a dashboard in Excel. I used Pivot Tables, slicers, and pretty colors of the data to show total units, historical purchases, predictability of new geographical areas, and trends on when assets were to be purchased. I made my presentation to management. Everyone was impressed with what the data showed. Management ignored my recommendations. We still have stock issues, and all my forecasting is proving accurate. Real-world reality – “You can lead a horse to water…..”

Many emergency departments throughout the country and around the world are experiencing long wait times. At our hospital, we are no exception. In order to resolve the problem, we need to identify the barriers. One of the challenges is having the right amount of staff and resources when the department experiences a surge of patients. If we could better predict hourly, daily, and monthly volumes based on historical data, the department may be able to balance staffing and resources.

To help solve this problem, I would need to utilize all the tools available to me, including R, SQL, CCL, PowerBI, Qlik, and real-time data feeds, to create a predictive model that could predict accurate volumes.

I would need to utilize historical weather data, including temperature and conditions such as sun, rain, or snow. I would also need to use moon phase data with daily percentages (to prove or disprove the full moon theory.) I would add calendar events around holidays, school schedules, report cards, end-of grades, government subsidy checks, and other community-type special events that draw large crowds. Most importantly, I need to add historical clinical data that includes dates, times, volumes, and reason for the visit to calculate recommended staffing patterns. I would also add historical epidemic data such as flu and COVID cases to help identify seasonal trends.

I would create a predictive analytic dashboard with historical data and a future forecast that is updated every 30-60 minutes. The combined data would help identify patterns and volumes for a specific time. This would allow the ED management team to see the daily forecast and predictions for the next seven and 30 days. The visualizations below are examples of some of the data and dashboards we currently have around ED volumes.