Create financial forecasts with generative AI: Trend analysis

- [Instructor] In our third and final case study, let's look at two approaches for forecasting Apple's revenue, trend analysis and predictive analytics. For both approaches, we'll be using historical quarterly data from Apple's financial statements as our time series basis to forecast revenue for the next 12 quarters. This kind of analysis used to be time-consuming using traditional tools, but today, you'll see how much time we can save using generative AI. One approach to forecasting is simple linear regression. Here, we fit a line to the historical revenue and extend that line into the future. Let's take a look. In this chart, you can see the straight line typical of linear regression. Historical revenue points are shown with the model extending the trend forward 12 future quarters, and that's a great chart, and linear regression is a great method, but we all know that real forecasts don't follow a perfectly straight line. A more complex approach is using SARIMA or Seasonal Autoregressive Integrated Moving Average, which captures different aspects of the time series, such as seasonality and trends. For those unfamiliar with SARIMA, it's a time series analysis method that models future data based on past values, differences between consecutive data points, and a moving average component to capture patterns and trends in the data. Let's check it out. Using SARIMA, we now see a forecast that captures the fluctuations over time. Rather than a single trend line, this forecast reflects potential seasonal effects and variability in the future revenue. So you can see with Apple that the fourth quarter is definitely the strongest quarter of the year, and our forecast shows that. Now let's look at the two side by side. So now we see the Seasonal Autoregressive Integrated Moving Average and linear regression in the same chart. Let's ask ChatGPT which approach would be better for our forecast? This is great. Just like a good financial analyst, ChatGPT has summarized the benefits and drawbacks of each method and made a recommendation based on the data. Both methods have their merits. Simple regression is easier to understand, but may miss seasonal patterns. SARIMA, while more complex, can offer a more realistic forecast by accounting for variations in trends. But no model is complete without domain knowledge. While data science techniques provide a quantitative foundation, it's crucial to overlay these models with qualitative insights. Understanding industry trends, product cycles, and macroeconomic conditions can help refine the forecast even further. By combining data science techniques with domain expertise, we create a more accurate and actionable forecast. This not only predicts revenue, but helps identify drivers and potential risks. Generative AI democratizes access to these powerful forecasting tools, allowing finance and strategy teams to be faster and more informed than ever before.