Classification
Format
Alexander Liss
Classification is a basic form of
scientific study, it is a simplest model. To provide Classification, one needs to
outline a set of Objects of Classification, to select a set of Characteristics
of the Objects and to describe how these Characteristics are “measured” in an Object
from the set. Note that the “measurement” is often limited to assertion that a
particular Characteristic is important in the object.
Each Characteristic from such
Classification has a set of possible Values, for example one could assert that
a Characteristic C is important in the Object O and it is of a type T1 (one
from a set of possible types T1,T2 and T3), or one
could assert that this characteristic is not important, when one analyses this
object (N/A). A set of possible Values of such Characteristic is: T1, T2, T3
and N/A (characteristic is not important).
When one combines Characteristics, not all
combinations of their Values make sense for a given set of Objects. Hence, in the
Classification one has to define, which such combinations are a part of the
Classification – a Class.
The description of combinations of Values
of Characteristics excluded from Classifications is a part of Classification.
This part contains important information about the Objects and selected
Characteristics.
All possible combinations of possible
Values of Characteristics form a set of Classes. An Object from the set of Objects
could be analyzed and all appropriate Classes from the set of Classes could be
selected to represent the Object.
Objects, which have similar representation
in the set of Classes (have similar sub-sets of Classes representing them), are
similar. This is used in analysis of new objects and in education.
Characteristics are often organized in a generic
hierarchical form. A group of characteristics could be presented as variants of
specification of another more generic Characteristic. In addition, the same Characteristic
could belong to different groups and hence present specifications of different
generic Characteristics.
When one analyses an object, one selects a Hierarchy
of Characteristics, takes the top of it and asks is this Characteristic
important for description of the object. When it is important, one decides
should be this Characteristic alone be used for Classification or it is useful
to proceed with Characteristics-specifications. When this is the Characteristic
is used without its further specifications, a Value is assigned to the Object.
Otherwise, one looks at the next level of Characteristics and ask the same
questions for each Characteristic-specification and so on. This procedure is repeated
for all Hierarchies. If there are some conflicts in assignment of Values of
Characteristics, because some Characteristics belong to different Hierarchies
and could be measured differently, when they are applied in different points of
the process, then these conflicts are resolved. In the end, a Class is assigned
to the object.
A Distributed Classification is divided in relatively
independent Units, where each Unit is maintained separately. In the Distributed
Classification, instead of descriptions (of Characteristics, Values, Hierarchical
trees, definition of Classes, etc.) references are used. Links are assigned to
these references and links are updated as needed. Using these links, a proper
description could be imported from a remote Unit.
A set of Objects of Classification for a
Unit is a subset of Objects of Classification of combined Classification.
As any model, the Classification evolves to reflect
growing knowledge in the area and expansion of the set of objects, which it
covers. Sometimes, a new Characteristic is added or a new possible Value for an
existing Characteristic is added. These additions are easy to maintain.
However, sometimes one subset of Characteristics
needs to be replaced with another. In such case, a new Version of the
Classification is introduced. In the Distributed Classification, Units have Versions
and a Version of the Classification is a set of Versions of Units.
Groups of Classes are defined through defining
subsets of Characteristics and for each Characteristic from that subset a
subset of Values. A Class belongs to such Group of Classes, when its definition
does not contradict the definition of the Group (it could have specifications,
which the Group does not have, but it cannot have specifications outside the
bounds set by the Group definition).
Groups of Classes are defined using Hierarchies of
Characteristics in a procedure similar to finding a Class of an Object.
When a Group of Classes is defined, the
Classification could be searched and a list of Classes belonging to this group
could be compiled.
As any meaningful scientific work,
designing a useful Classification is difficult, but when it exists, it provides
substantial support, when one tries to find an applicable Object for a task or
tries to figure out limits of applicability of a given Object.
Following summarizes what is included in
the Classification:
·
Description
of Units of Classification (including references and links)
·
A
Version of Classification (a list of Versions of Units)
·
A
set of Characteristics and their possible Values (each Unit separately)
·
A
set of Hierarchical trees on the set of Characteristics (using references to
Units, when needed)
·
A
set of Objects of Classification
·
A
method of assigning a Value of Characteristic for each Object (in a Unit, where
the Characteristic is defined)
·
A
set of combinations of Values of Characteristics excluded from Classification (each
Unit separately)
·
A
(partial) list of Classes with Values of Characteristics, which define them and
examples of Objects, which belong to them
Following provides elements of Formal
Description of the Classification, which allow automated searches through
Classification and sharing of information between Units. All is encoded in XML;
names of nodes are in brackets.
Each Classification Unit provides its name
[name], information about its Version [version], and description of other Units
[units].
The description of a remote Unit [unit] includes:
Description of each Characteristic [factor]
includes its name [name], description [description], description of ways to
assign a Value to an Object of Classification according to this Characteristic
[measurement] and a list of possible Values [values], where each Value [value]
has a name [name] and description [description].
Description of each hierarchy [hierarchy]
contains its name [name] and a description [description] of a tree, which is
convenient to describe using XML. It is described with nodes [nodes], where
each node [node] contains a name of Characteristic and a list of child nodes.
Each node could contain description [description] describing how its
Characteristic relates to the Characteristic of the parent node.
This part is a free form description of
Objects of Classification
It is a list of Group of Classes of Objects,
where each element of the list [group] contains description of the group
[description] and a list [limits] of combinations [limit] of Characteristic’s
name [name] and ranges of Values [ranges], where an element of this list
[range] contains either a Value or a
range of Values of given Characteristic.
A Class [class] (an element of the list of
Classes) contains a list of Characteristics and their Values, which define it
[definition], where each element of this list [unit] contains a name (reference
name, in case of remotely defined Characteristic) [name] and a Value [value].
In addition, it contains a URL link [link] to the Class description, examples
of Objects and their use. These descriptions are perpetually updated; hence
they are kept in a separate place.
A request is description of a Group of
Objects of Classification. The response to such request is a list of Classes.
Both are encoded with XML.
It is a list of combinations [limits] of
Characteristic’s name [name] and ranges of Values [ranges], where an element of
this list [range] contains either a Value or a range of Values of given
Characteristic.
It is a list of Classes, where each element
of the list [class] contains a list of Characteristics and their Values, which
define it [definition], where each element [unit] contains a name (reference
name, in case of remotely defined Characteristic) [name] and a Value [value].
In addition, it contains a URL link [link] to the Class description.
Classes are associated with Objects 9lists
of references, for example). Secondary Query retrieves lists of these Objects.
Querying of the Classification is supported
with GUI.
Entire Classification is presented in a way
reminiscent of a file explorer, where nodes of Hierarchies (Characteristics)
correspond to directories (with the difference that there are duplicate nodes
in such presentation). These nodes could be expended or collapsed with a mouse
click. Each node “directory” contains a “file”, clicking on which opens GIU
module describing available Values of the Characteristic and providing ability
to select a subset of Values.
When all selections are done, the check of
selection consistency could be initiated and query into the Classification
could be submitted.
The result is presented as an HTML page
with links to descriptions of Classes.
Broken links to other Units lead simply to
large number of truncated hierarchies.
If a Secondary Query is involved, then it
is executed using a list of Classes and presented according to the nature of
the presentation of lists of Objects.
Examples
1. Classification Unit of Properties
of Parameters of a Model and Results of their Measurement
<head>
<name>
properties of parameters and results
of their measurement
<name>
<version>
1
<version>
</head>
<factors>
<factor>
<name>
type of parameter
</name>
<description>
In The Art of Optimization (see this site)
parameters of models are divided into three types, the type of parameters used
affects what kind of conclusion could be made using the model
</description>
<measurement>
A parameter of the model is assigned its type
according to procedure described in The Art of Optimization
</measurement>
<values>
<value>
<name>
technical
</name>
<description>
Parameters like length, weight, time, fuel
efficiency, etc.
</description>
</value>
<value>
<name>
money based
</name>
Parameters like cost, price, interest rate, etc.
<description>
</description>
</value>
<value>
<name>
conventional
</name>
<description>
Parameters like “acceptance of risk”, “warmth of
color, etc.
</description>
</value>
</values>
</factor>
<factor>
<name>
type of result of measurement
</name>
<description>
The type defines what kind of operations could be
performed with results of measurement, see The Art of Optimization
</description>
<measurement>
Type of the results of measurement of a parameter is
a well known characteristic of measurement; it is specified together with
measurement procedure
</measurement>
<values>
<value>
<name>
order
</name>
<description>
Only an order of results of measurement is
meaningful, (hardness)
</description>
</value>
<value>
<name>
difference
</name>
<description>
Difference between Values (and, hence, order of
values) is meaningful (temperature)
</description>
</value>
<value>
<name>
ratio
</name>
<description>
Ratio of Values (and, hence, difference between
Values and order of Values) is meaningful (mass, volume)
</description>
</value>
</values>
</factor>
<factor>
<name>
sources of information
</name>
<description>
This is an indirect way of assessing quality of data
used to arrive to results of measurements of parameters of the model
</description>
</factor>
</factors>
<excluded>
<group>
<description>
Results of measurement of money based parameters are
of type “ratio”
</description>
<limits>
<limit>
<name>
type of parameter
</name>
<ranges>
<range>
money based
</range>
</ranges>
</limit>
<limit>
<name>
type of result of measurement
</name>
<ranges>
<range>
order, difference
</range>
</ranges>
</limit>
</limits>
</group>
</excluded>
<classes>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
technical
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
order
</value>
</unit>
</definition>
</class>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
technical
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
difference
</value>
</unit>
</definition>
</class>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
technical
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
ratio
</value>
</unit>
</definition>
</class>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
money based
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
ratio
</value>
</unit>
</definition>
</class>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
conventional
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
order
</value>
</unit>
</definition>
</class>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
conventional
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
difference
</value>
</unit>
</definition>
</class>
<class>
<definition>
<unit>
<name>
type of parameter
</name>
<value>
conventional
</value>
</unit>
<unit>
<name>
type of result of measurement
</name>
<value>
ratio
</value>
</unit>
</definition>
</class>
</classes>
2. Description of Hierarchy
of Characteristics
<hierarchies>
<hierarchy>
<name>
description
of the object’s functioning
</name>
<description>
Functioning of objects of optimization could be
described in more or less formal and detailed way or could be inferred from
experimentations with the object (see The Art of Optimization)
</description>
<node>
<name>
description
of the object’s functioning
</name>
<node>
<name>
structure of the
object
<name>
<node>
<name>
number of elements
<name>
</node>
<node>
<name>
type of connections between elements
<name>
</node>
</node>
<node>
<name>
source of information
<name>
</node>
<node>
<name>
macro models used
<name>
<node>
<name>
type of theory used to make the
model
<name>
</node>
</node>
</node>
</hierarchy>
</hierarchies>