Speaker 1: My smartwatch tracks how much sleep I get each night. If I'm feeling curious, I can look on my phone and see my nightly slumber plotted on a graph. It might look something like this. And on the graph, on the y-axis, we have the hours of sleep, and then on the x-axis, we have days. And this is an example of a time series. And what a time series is, is data of the same entity, like my sleep hours, collected at regular intervals, like over days. And when we have time series, we can perform a time series analysis. And this is where we analyze the timestamp data to extract meaningful insights and predictions about the future. And while it's super useful to forecast that I am going to probably get, like, seven hours shut-eye tonight based on the data, time series analysis plays a significant role in helping organizations drive better business decisions. So, for example, using time series analysis, a retailer can use this functionality to predict future sales and optimize their inventory levels. Conversely, if you're into purchasing, a purchaser can use time series analysis to predict commodity prices and make informed purchasing decisions. And then in fields like agriculture, we can use time series analysis to predict weather patterns, influencing decisions on harvesting and when to plant. So, let's first of all introduce, number one, the components of time series analysis. And then, number two, we're going to take a look at some of the forecasting models for performing time series analysis. And then, number three, we're going to talk about how to implement some of this stuff. Okay, now, let's talk about the components first of all. And one component is called trend. Now, this component refers to the overall direction of the data over time, whether it's increasing, whether it's decreasing, perhaps it's staying the same. So, you can think of it like a line on the graph that's either going up or going down or staying flat. That's the first component. The second one, seasonality. Now, this component is a repeating pattern of data over a set period of time, like the way that retail sales spike during the holiday season. So, we might see a spike and then a bit lower, the spike is back and it keeps repeating like that. That's seasonality. Third component, that's cycle. And cycle refers to repeating but non-seasonal patterns in the data. So, these might be economic booms and busts that happen over several years or maybe even decades. So, it's a much smoother curve. And then lastly, there is variation. And variation refers to the unpredictable ups and downs in the data that cannot be explained by these other components. And this component is also known as irregularity or noise and well, it looks like maybe that. Yeah, very difficult to pick out the trend. So, those are some of the components of time series but let's talk about the forecasting models that we can use to perform some analysis. And there are several popular forecasting models out there. One of the most well-known is called the ARIMA model. Now, ARIMA, that stands for auto-regressive integrated moving average. And the model is made up of three components. So, there's the AR part, that's the auto-regressive component and that looks at how past values affect future values. Then there's the I for integrated or differencing component and that accounts for trends and seasonality. And then there is the MA component, that's the moving average component and that smooths out the noise by removing non-deterministic or random movements from a time series. So, that's ARIMA. Another pretty popular one you'll often see is called exponential smoothing. And exponential smoothing model is used to forecast time series data that doesn't have a clear trend or seasonality. So, it doesn't fit into these kind of areas. And this model works by smoothing out the data by giving more weight to recent values and less weight to older values. And there are many other forecasting models out there and the right one to use, of course, depends on the data you're working with and the specific problem you're trying to solve. Okay, so let's finally talk a little bit about implementation. How do we implement this? There are several software packages out there that can help you perform time series analysis and forecasting such as those with R and Python and MATLAB. So, if we just focus in on Python for a moment, two of the most popular libraries for time series analysis in Python, firstly, pandas, and secondly, a library called matplotlib. With pandas, you can easily import, manipulate, and analyze the time series data and it can handle things like missing values, aggregate data, and perform statistical analysis on the data. matplotlib is a library that can help you visualize the time series data. You can create line charts or scatter plots and heat maps. Using these libraries, you can perform a wide range of time series analysis tasks like data cleaning, exploratory data analysis, and modeling. You can use pandas to pre-process your time series data and then use matplotlib to visualize the trends and seasonalities in that data. By understanding the components of a time series and then choosing the right forecasting model, you can make more informed decisions and gain a competitive advantage. So, look, whether you're a data analyst or a business owner or just a curious sleeper, take advantage of the power of time series analysis and get a glimpse into what the future may hold.

Speaker 2: If you have any questions, please drop us a line below. And if you want to see more videos like this in the future, please like and subscribe. Thanks for watching. www.microsoft.com