Time Series Data Features:
There typically exists three significant Data Features in a time series data, namely Trend, Seasonality and Irregularity. We will briefly discuss these three features with a sample example
Trend: Trend is often described as varying mean over time and is often present in the time series data. In our sample product sales data as depicted in the figure below, we observed the positive trend up until March 12th and then after March 12th, we observed the negative trend in the sales i.e. we observed that the sales of the product gradually increased from Jan 1st till Mach 12th and then after 12th March the product A’s sales started gradually decreasing.
Seasonality: Seasonality is often described as the variations at specific time frames. For example, people might have tendency to buy more cars month because of pay increment or festivals.
In our particular sample product sales case, we observed that the sales of product appear to follow certain seasonal patterns as shown in the Figure 2 below. The seasonality has been highlighted in 4 red circles, with each circle representing a pattern, in the figure.
Irregularity: In our sample data, we observed the residual trend i.e. irregularity trend. The irregularity or residual trend in our data does not follow a particular pattern. The irregularity pattern is shown in the yellow circles in the Figure 2 below.
We observed that the residual pattern follows less fluctuations in the early time series period, as depicted in the first yellow circle in the Figure 2. In the mid-time series, as depicted in the second yellow circle the irregular trends are larger in magnitude / variations but for shorter period. At the end time of the time series as represented by the third circle we observed that the residual errors pattern has changed again, with more significant increase in both the intensity and the period length. This irregular trend increase the complexity for making prediction for our data more difficult.