Quick Insights for Multi Time-Series Data

Intended audience: END-USERS DATA SCIENCE DEVELOPERS

AO Easy Answers: 4.3

Overview

Quick Insights can be generated at runtime by the user from Apps on the Easy Answers Results page, or from a scheduled task created by the Solution Developer using the Insight Composer in the AO Platform. This topic covers the Quick Insights relating to Multi Time-Series Data.

Multi Time-Series Quick Insights

Multivariate Trend Detection (Bayesian) - The Multivariate Trend Detection generates one Global Trend Line Insight for all series as well as one Individual Trend Line Insight and Trend Analysis for the Time Durations selected for each time series.
Multivariate Change Point Detection (Bayesian) - The Multivariate Change Point Detection generates one Individual Change Point Insight for the Time Durations selected for each time series.
Correlation - Time-Series - The Multivariate Global Insight generates a Correlation Heatmap Insight for all series.

Configuring Multi Time-Series Quick Insights

The Multi Time-Series Quick Insights have a number of Common Properties as well as a few additional unique properties per Quick Insight. When this Quick Insight is selected, the user can configure all the properties, or leave the default configuration as is and just decide on which individual Quick Insights should be added, eg. Outliers, Change Points, Trend Lines, and/or Correlation.

Common Properties

Label	UI	Default	Description
Select Series	Tags field w/dropdown		Dropdown with multi-select - The user can select two or more field properties as input to the Quick Insight models.
Model Size	Dropdown	Default (2500)	Manage the model’s size. Larger models provide reliable interpretations at the cost of longer insight run times. Smaller models produce quicker results but lack the inclusiveness of larger counterparts. Default - Samples: 2500 Custom - enter a value in the Number of Trials field below.
Number of Trials	Number Field	2500	Enter a number for the number of Trial Samples. This field is only visible if the user has selected Custom in the Model Size field above.
Type of Data	Dropdown	Data is Normal	Select the Type of Data from either of the following options: Data is Normal - data is expected to have a normal value distribution. Data has Outliers - data is expected to have a distribution with some outliers.
Analyze By	Dropdown	Day	Time durations for which the data will be analyzed. Multi-select (tags) for one or more time duration values, including: Hour Day Week Month Quarter Year
Show Analysis for Periodic Estimation	ON/OFF Toggles		One or more On/Off Toggles will show depending on the selection for Analyze By above.
Special Day Insight	ON/OFF Toggle	ON	Will generate additional insights relating specifically to configured Special Days for the Ontology. This toggle is only available if Day is selected for Analyze By above.
Number of Candidate Change Points	Number field	5	Defines how many key trend shifts the system considers when analyzing business data. A higher number captures more variations, but may lead to unnecessary complexity. Example: Tracking daily fluctuations in stock prices. A lower number keeps insights simpler, focusing only on major changes. Example: Identifying major sales peaks in a quarterly revenue report.
How to Handle Missing Values	Dropdown	Replace with previous value	Select how the Insight model shall handle records with no values Replace with predicted value Replace with average value Replace with previous value Replace with zero

Summary of Quick Insights

	Multi Time-Series	Correlation - Time-Series
Number of Series	Should have at least 2 time series	Should have at least 2 series
Type of Data	Continuous or Discrete Data	Continuous or Discrete Data
Charts Examples	Time-Series Chart	Scatter Plot / Time-Series
Examples	Time series of multiple stocks' data will give insights for trend, seasonality, outlier, and changepoint detection, global trend for all series, and correlation for each pair of series.	Time series of 2 stocks data will give insights into positive or negative correlation.

Contact App Orchid | Disclaimer