Nonnumerical data

In the logical-structural approach, an attempt is made to include procedures that model the researcher s intentions. This is done by introducing some interactive (dialogue) regimes in the work of the researcher with the computer. Finn and co-workers created a computer version of the so-called verisimilar reasoning (in the style of J. C. Mell) worked out for automatic creation of hypotheses based on a cause effect analysis of nonnumerical data under the conditions of incomplete information. The purpose of verisimilar conclusions is to reveal similarities and distinctions in the structural graphs of different molecules in the series under consideration, and these similarities and distinctions should correlate with the molecular properties. 

All the preceding discussion has concerned continuous numerical data. Not all data is of this form, and we now discuss the encoding of nonnumerical data. Logical variables have one of two values that are encoded as zero and one. An example is the presence or absence (a yes or no) of a feature. Typically the presence (yes) is encoded as a 1 and the absence (no) as a 0. Categorical data should be encoded in some variation of a one-of-N code (N equals the number of categories) or a thermometer code. 

Yager 94 Piece of data Neural Networks Yes Can Handle nonnumeric data, developed a function that transforms a group of data into a sir le data point... 

A sequence of computational instructions which yield a solution to a specific problem. Algorithms can be numerical or analytical, or can operate on nonnumerical data such as strings. One important aspect when selecting algorithms for larger problems is efficiency, especially the scaling of execution time with input data size, which is typically represented by an expression such as logn, or /iP-compiete worst case,... 

A general concern in data mining is the representation of objects. Molecules, text documents, images, nucleic acid, or protein sequences all represent nonnumerical objects. However, all data mining methods require the transformation of objects into an algebraic, i.e., numerical, representation. 

The data are the unevaluated, independent entries that can be either numeric or nonnumeric e.g., alphabetic or symbolic. Information is ordered data that are retrieved according to a user s need. If raw data are to be effectively processed to yield information, they must first be organized logically into files. In compnter files, a field is the smallest unit of data, which contains a single fact. A set of related fields is gronped as a record, which contains all the facts about an item in the table, and a collection of records of the same type is called a file, which contains aU facts about a topic in the table. Databases are electronic filing cabinets that serve as a convenient and efficient means of storing vast amounts of information. 

Dummy Variable n Also known as an indicator variable is a variable generated from a non-quantitative or nonnumeric categorical feature of the data in order to associate a quantitative value with the feature so that its effects can be used in regression analysis. As a standard, the variable is given a range of zero to one. A dummy variable is often used to indicate the presence of a feature such as left-handedness. Multiple dummy variables can be used to represent independent but related categories such as the days of the week. 

The risk assessment can be processed in a qualitative or quantitative framework. The qualitative risk assessment will be based on methods, principles, or rules for assessing risk based on nonnumerical categories or levels. An example would be where the safety manager assigns categorization of risks as low, medium, or high based on known data or assumptions regarding the risk environment. 

---