Time-series II (seasonality and forecast)

In this blog post, I will introduce two new time-series data analysis features of HEARTCOUNT.

Previous blog about our time-series feature can be accessed here

if you want a hands-on experience:
- visit our free demo site: https://www.heartcount.io/play and select a sample "superstore" dataset from top menu
- or upload your own dataset which contains at least one date/time column.

1. Seasonality

A repeating pattern within each year is known as seasonality. Business KPI such as sales are greatly affected by seasonal factors such as the time of the year or the day of the week. Hence, quantitatively understanding seasonal pattern is important.

• if you visualize sales(sum) over time(order date), it’s quite difficult to grasp the seasonal patterns.

[time-series visualization with interval set as week]

• HEARTCOUNT automatically creates seasonality(periodicity) variables for every date/time variable as shown below.

• so, if you’re interested in finding “days of week” sales pattern:

[subgroup is set a x-axis(=days of week) and bar chart icon is selected)

• or, compare the seasonal(days of week) sales pattern per each category:

[stacked bar icon is selected here]

2. Forecast

HEARTCOUNT uses AR(AutoRegression) models, which is a time series model that uses observations from previous time steps as input to a regression equation to predict the value at the next time step.

There is no lack of literature on AR, so I will just brief about “how-to use forecast feature”.

• whenever X-axis variable is set as date/time variable, there is “forecast” icon displayed as follows:

• two forecast lines, one based on least square method, the other based on Max Entropy, are drawn as prediction range as follows: