Hello. In this video, we'll use our definition of data science to model various examples of data science. Recall our definition of data science. Data science is an interdisciplinary field made up of domain knowledge, applied statistics, computer science, and machine learning with the goal of using data within a scientific process to develop and apply knowledge in a specific domain. This definition can be used to model almost any data science project. But let's take a look at our supply chain management example. We can represent this visually and use the framework to design our data science project. Let's start with a data-centric scientific process. First, we need to determine our question. Given our domain knowledge and customer survey data, we see that our stores have been having out-of-stock problems recently. As a result, we want to know if we can reduce the number of out-of-stock items in our store. Next, we need to determine our hypothesis set. Our null hypothesis could be that the number of out-of-stock items in the store remains constant after our treatment and the alternative hypothesis and in this case, the one we're hoping for is that the number of out-of-stock items in the store decreases after our treatment. These hypotheses represent opposite answers to our original question. So we'll conclude with only one. Next, we need to carry out our experiment or analysis. Given our technical and domain knowledge, we know that using a predictive model to predict when an item is likely to go out of stock could be a good way to reduce the number of out-of-stock items. It would provide our ordering system with an estimate of when to order more of each particular item in the store. We'll need to collect historical data on item inventory and items sales for our model to learn to use the relationship between the two. We can also use any other data we might think has a relationship with item inventory or items sales. Once we have all the data we'd like, we can go ahead and build a model to predict when an item is likely to go out of stock. Then, we need to evaluate the model. We might see according to pass data that it's predictions are pretty accurate and will reduce the number of out-of-stock items if used correctly. At this point, if it's possible, it would be best to release these predictions to a small sample of stores, as in market tests to see if the model truly does reject the null hypothesis by reducing the number of out-of-stock items in the store in a real-world setting, rather than just in a original model evaluation phase. Once we get our definitive result from a real-world setting, we can communicate and deliver our results. If we conclude that the model did reduce the number of out-of-stock items, we will likely want to describe the model and process to key stakeholders in some type of presentation. Next, we should set up some way to deliver these predictions to those responsible for ordinary new items, whether that's an automated ordering system or storm integers. Finally, it would be wise to set up a real-time monitoring dashboard to understand how our model is performing at all times. After a while, it might need to be refreshed due to changing relationships between inventory and sales. On the other hand, if we conclude that a model did not reduce the number of out-of-stock items, we still need to create some type of presentation describing why it did not work as intended and any possible recommendations for changes that can make the project more successful. It can be really helpful to fill out the decisions, the need to be made for a specific data science problem at each of these steps of the scientific process. But we also need to identify which technical skills will be necessary along the way. Throughout this process, we will need to use all of the skills in our data science toolkit, domain knowledge, data literacy, data manipulation and transformation, statistics, machine learning, model evaluation, programming, dashboarding, and working with REST APIs and real-time streaming. It might not be reasonable to expect that a junior level data scientists have all of these skills necessary to complete and deliver a project, but outlining which skills will be necessary at different stages can help entire data team's plan where other data practitioners like data engineers or machine learning engineers might need to be involved with this project. Well, this example looked at supply chain management. You can complete this process for just about any data science project. A few examples include targeted advertisements toward customers that are more likely to buy a product, predicting when a stock is likely to rise or fall in a financial market, determining whether a new vaccine is effective in preventing the spread of a disease and predicting which candidate might be likely to win an election, among many others across many other industries. For each of these projects, it can be helpful to design the data science process ahead of time by identifying the decisions and actions that need to be taken at each step, and identifying the skills necessary to complete those actions. At this point, we hope you have a solid understanding of what data science is, including its purpose, the skills required to perform data science, and the process that should be followed. You should also be comfortable with the idea of framing real-world problems in this data science process. To close out this lesson, you will complete an activity where you will design your own data science project using the process exemplified in this lesson. Thanks for joining us.
