TimeTransf Module

List of windows:
TimeTransf module
Main window

Task detail
View results

Graphs & Analysis
Data view
Graph view
Frequency analysis

Group of tasks
Custom items select
TimeTransf module

TimeTransf module is used to preprocess data containing time series to be used with LISp-Miner. This is done by computing characteristics for each time series. You define your own characteristics depending on which information you need to gain from time series.

Data model

Your database must contain at least one event matrix (table with time series).

Each event matrix must include:
  • event id - a unique primary key for each event
  • object id - common id of events within one time series
  • date or time value - explains to us when the particular event happened
  • value - the actual value of event of the particular time series
  • symbol - optional value which might be used to separate certain groups of events or to filter some events

Object matrix associate event matrices through ObjectIDs. You can compute characteristics from all event matrices of particular ObjectID simultaneously for all time series.
If you have only one Event Matrix you don't need to use Object Matrix. Set "Field for key items" in Task detail window to the ObjectId of your Event Matrix otherwise set it to the ObjectID of Object Matrix.

Scheme of data model. The left part represents original data. The right part shows the matrix with computed results as your characteristics. The arrow represents computing transformation.

After the computing process is finished you will find Characteristics Matrix with the results in your original database next to your event matrices. This matrix is ready to be used with LISp-Miner.
In the Characteristics Matrix there is one row for each object. First column contains ObjectID's. Next columns contain result values for particular characteristics. The number of columns depends on how many characteristics you have defined.

For better understanding of this problem look closer at the following example using financial data.

Make sure that database columns used by date/time and symbol variables are indexed. Computing process speed will increase noticeably.

TimeTransf module needs following files to work properly:

  • timetransf.exe
  • borlandmm.dll
  • cc3250mt.dll
  • cg32.dll