Nonstructural Data Processing of Expert System for Evaluation of Location Region Touristic Cluster

MARTYSHENKO N S

(Vladivostok State University of Economics and Service，Vladivostok 690990，Russia)

One of the key problems in fundamental and applied science in the sphere of tourism is the development of effective patterns for forecast and control of tourist flows.The main difficulty of its elaboration and implementation is the lack of information on touristic product consumption in the region and tendencies in development of consumer's needs in recreational，leisure and other activities.This information can be collected during questionnaire surveys of touristic product consumers.

More than 8 years ago authors started working on monitoring of touristic product consumers behaviour in Primorskii kray.During this period a number of surveys were carried out［1］.

As it was cleared up the range of main features of touristic product consumption process cannot be defined using regular structural questions.The experience showed that the person answers the quality and comparative questionsbetterthan quantity ones.

Therefore our survey data contains mostly nonnumeric information due to using different measuring scales in polls.The purpose of this method is to obtain the correct information.The researcher always must look for the compromise between desirable information and the one that he collects during surveys.When composing the questionnaire we had to use different forms of open questions.

Open or nonstructural questions are the most difficult for computer processing of survey data.Unlike close questions，nonstructural questions don't contain prompts，don't“impose”one of the variants and collect unformed opinion.

The problem of the computer analysis of the qualitative data involves a considerable quantity of researchers all over the world.Definition of a problem of the computer analysis of the qualitative data is given in work［2］.In the Russian scientists specialising in area of processing of the qualitative information it is possible to name Davydova A.A.［3］，Kanygina G.V.［4］.

Known computer means of processing of the qualitative data，such as NVIO［5］are difficult enough for the majority of experts in marketing.Therefore was attempt of creation of simple computer means which is undertaken could be well for development to the wide range of researchers.

In the present work we suggest to overview some element of this technology，which let us transfer nonstructural information into structural (typical)one.Then we offer to study the examples of typology construction and its use in analysis of location development of touristic region.

1 Key Elements of Open Questions Computer Technology

As a rule the questionnaire data are compiled into“subject-quality”table.It is easy to place this table at EXCEL page.Text data on an open question is composed into one column.Also we take into account that the answers may be numerous.For example，the answer to the question“What do you prefer to do during your rest at the ocean，except sunbathing and swimming?”may be“playing volleyball，sightseeing，fishing”and so on.We name the feature in the“subject-quality”table as a component.

The answers in compound feature contain several simple statements.Simple statements are separated with a sign“;”or“，”.Complicated cases contain the whole sentences.In simple cases the answer has only one statement.So we do not limit respondent in his answers.

Then we define what we have at the outcome of developed information technology.Let's start from the simple statements which are special case in compound feature.The respondents are expected to give different answers.Still experience shows numerous similar answers let alone spills and spelling mistakes.Range of answers that are different in sense but not in form is limited.The sample at the amount of 700 questionnaires shows from 30 to 50 different answers that can be interpreted as a meaning of the feature measured by nominal scale.The list ofvariants does not change if the sample volume is increased.

We use the technique of“typification”for open questions processing.This technique is the change of original simple statement to close or resumptive statement.The specific table-“list of feature meaning”-is composed for typification.During this process compound feature is split to simple ones.One of the table columns contains all unique meanings of the original feature and the column of their frequency.Typification is applied for table“list of feature meaning”，not for“subject-quality”table.At first the simple cases are to be processed，for example different spelling of the word or different word orders.The most precise and correct statement is chosen among similar phrases and copied to the cells of“list of feature meaning”table.We reduce the number of lines in this table with the help of the change of the unique statement to that one from the list of meaning.The next step is“compaction”in the feature meanings.After reduction“the list of feature meaning”table acquires visual and demonstrative view.

Later on we process the complicated situations.We find infrequent but similar in its meaning groups of statements in“list of feature meaning”table.Then the researcher chooses one common statement for this group or composes the new one with the same sense or theme.

For example，answering the question“What do you prefer to do during your rest at the ocean，except sunbathing and swimming?”respondents reply in the following way“riding jet-ski，hangglider running，water skiing，diving from the rocks，rock climbing”.But these statements are infrequent so we refer them to resumptive category“extreme”.In such a manners the meaning stays the same.

To avoid the information loss we decided to specify common statements in the brackets.For example in the case mentioned above we changed the original statement to:“extreme”(riding jetski)，“extreme”(hang-glider running)， “extreme”(water skiing)，“extreme”(diving from the rocks)，“extreme”(rock climbing).

The manner of answer is more important than the precise meaning so as the manner defines the personality pattern.The original“list of feature meaning”table has several thousands meanings.After typication the number of them reduces to 300 taking into account specifications.This contracted table is the first level of typification.

The given new feature still has too many meanings to analyze.Therefore the second level implements additional processing.At this stage we exclude specifications and form one more column of the“list of feature meaning”table which is called underclass with 30～50 unique statements.

It is still difficult to measure 30～50 variants by the nominal scale.For this reason the researcher has to group given answers considering them as characteristics of non-crossing classes，types and topics according to the feature meaning and the goal of typification.In our case the personal pattern determination is more suitable.Integration of simple statements into classes is the third level of typification.The researcher names each class as it thinks fit according to the meaning of statements.

As a result after open questions processing we will have:

Three new features that are included in the original data table and can be processes further to get substantial output;

“List of feature meaning”table for using it in case of repeated questionnaire survey or data typification of different surveys within present research.

It is important to note that after compound features typification the compound features will be formed.Specific processing techniques are developed for its analysis.

Free-answer questions processing technology has one more important result that allows to lessen powerfully the time of data typification in case of repetition of surveys(process of monitoring).When replenishing the raw data table it is necessary to repeat the typification taking into account new information.In order to speed up the process of work a researcher can apply two types of dictionaries that are created for every identifier.The identifier contains free-answer question data:“Replacement dictionary”and“Cue words dictionary”.Such dictionaries are created for every separate qualitative character.Besides，another type of dictionaries is used.It works with different qualitative signs and also with different questionnaires and it is known as“Redundant information dictionary”.This type of dictionary is applied at the first stage of the qualitative text information process.It helps to delete or correct utterances，containing redundant and empty information.

All the dictionaries are kept in the same file Access.They represent knowledge base as the result of the researcher's work connected with utterance typification.

2 Development of Typologies of Tourist Product Consumers

One of the simplest examples of using the technology is typification of recreation areas visited by local population，for instance the population ofPrimoryeTerritory，in summer.The questionnaire“The research of beachfront waterbased recreation”contained two of such free-answer questions:

I often visit a beach:__________________ __________________(name of the bay，island or the nearest settlement);

I often visit recreation zone:____________ _______________(name of the nearest settlement，bay or island).

The first question touches on the overnight stay recreation，the second question touches on the recreation without staying overnight.Answering the questions respondents can inform about the several favorite beaches or recreation areas.The list of the possible recreation zones is rather long let alone the full list that cannot be made by the researcher himself.That is why in order to reveal the structure of recreation zone space distribution one can apply a free-answer question only.In practice this task is not as simple as it may seem.The practice shows that there are a lot of different ways of spelling of the places'names and the mistakes are often repeated at that.

Administrative district belonging can be chosen as a criterion of recreation areas classification.After processing unclassified data the information become arranged.In this case，initial information is classified without any loss.

Let us consider more complex situation.For example，in the questionnaire“The research of beachfront water-based recreation”we were interested in preferences of consumers in spending the time on the beach.As the people on the beach not only sunbathe and swim but do something else especially if they stay on the beach for more than one day.In order to know how people prefer to spend their time on the beach another question was included in the form:“What else do you like to do on the seashore except sunbathing and swimming:___________________”.

It turned out that range of interests are not so wide.After typification all the utterances were integrated into 8 groups:

1.Sportsmen:26%;

2.Inactive persons:19%;

3.Carried away by any activity:16%;

4.Gourmets:13%;

5.Lyric persons:12%;

6.Communicable persons:8%;

7.Sleepyheads:5%;

8.Mothers:1%.

This form was also used in order to research the negative expressions of the respondents who had a rest by the sea.To analyze the negative opinion the question“What darkened your recreation at the seaside:_____________”.The typification determined quite stable structure of negative consumer responses allocation.As the result of negative utterances grouping 9 groups were exposed:

1.Greens:53%;

2.Fastidious persons:12%;

3.Optimists:10%;

4.Unsociable persons:8%;

5.Urbanists:6%;

6.Intellectuals:5%;

7.Intolerant persons:2%;

8.Automobilists:2%;

9.Students:1%.

The method of computer typification also was applied in analysis of vacation pastime.

In the survey by questionnaire“The research of Primorsky region tourist potential and its prospects”the most complex form of free-answer questions was used.It is a special type of questionnaire that is based on the questions requiring the answers consisting of several sentences.For example“What conditions must a beach meet in order that you can visit it more often?”Even analyzing such a complex question the method of typification allowed to revealcertain groupsof Primorians'preferences.Fig.1 demonstrates frequency range of preferences distribution in accordance with integrated data.The figure's 1 diagram shows among others the main condition that a beach should meet in order that the locals can visit them.It is the improving of sanitary state of beaches(31%).

These numerical estimates are well-grounded.Regional authorities and business corporations should take them into consideration in order to provide consumer services taking into account variable tourists'demands in the different kinds of activity such as leisure activity，health improving，learning activity and others as the latter are the ba-sis and incentive for entrepreneurs to look for and implement new regional tourist products and types of tourism.

Fig.1 Grouping of respondents'answers about improving beach facilities

We examined three types of free-answer questions.First type of question supposes the answer consisting of one or several words，second type expects one or more than one simple phrases and the answer to the third type of question is supposed to contain complex sentences.

3 The Analysis of the Structures of Regional Touristic Complex Spatial Development on the Basis of Typologies

Any recreation resources are supposed to be tied to the certain place or to be determined by spatial location.In order to forecast and regulate touristic resources consumption it is necessary to learn how to describe the processes with the help of structured data.The range of the most important structured features is got as the result of typologies development.

The characteristics of pattern of consumption interact with each other and vary with time.The aim of strategic management is to direct the changes into the proper way.And it is necessary to develop a structural change analysis technology.

It appears that objects in their pure form cannot always be described by one structure feature.For example，some respondents'answers can be put into several groups.For example，when answering the question:“What darken your rest on the seashore?”，a respondent can answer:“ecological situation，public gathering，unequipped beach”i.e.Some simple expressions during the sign typification were put into different groups.In this case in the process of typification initial data were replaced by the names of the groups such as:“Greens”，“Unsociable persons”，“Urbanists”.Thusgroupscan coincide.However，some groups are close to each other，some are wide apart and the others are isolated completely.Let us consider a calculation principle of class intersection estimates.

For every compound answer，specified by the classes，combinations of couples of classes can be summarized.For example，if the separate answer was presented as:“S，I，I，S，M”(where the classes are determined by the letters)，the answer of the respondent can be put into three groups at the same time.Using this answer three combinations of groups can be compiled:“I，S”，“S，M”，“I，M”.Thus，taking into account possible class intersections intersection matrix can be arranged.Order of a matrix is К×К where К is the quantity of stood out class.The matrix is symmetrical with respect to diagonal.Matrix components are the sum of pair classes met in the bulk of sampling.To eliminate the influence of sample size and class size on the intersection estimates，array components are normalized by dividing the strings by the quantity of expression pairs according to classes.Diagonal elements describe isolation degree of separate groups.For example， classified data in relation to respondents'pastime preferences on the beach were used in creating the matrix presented in the table 1.

Table 1 Class intersection matrix

For demonstrating the class intersection it is convenient to graph the model(fig.2)

Fig.2 Class intersection graph

When graphing a threshold value of the intersection estimates is determined.In this case the graph will show only the most essential connections.The graph presented by fig.2 has a threshold value 0，1.Points of graphs named according to the first letters of their names.The graph demonstrates how it is necessary to combine range of provided service，that is to design regional tourist product for different groups of consumers.

4 Conclusion

(1)The report offers the technology of unstructured data processing，that will help to expand researchers opportunities in studying socio-economic phenomena and processes.

(2)The technique contains expert system that is based on the using of three types of computer dictionaries.With the help of the dictionaries researchers working in the same field，can exchange the results of their work connected with data classification.

(3)Providing information in the form of compound indicators allows to analyze cooperation of different structure consumer groups.

(4)The data processing technique offered was tested in sizable actual data levels processing［6－7］and can be recommended for wide circle of researchers conducting questionnaire survey in their practice.

［1］ Martyshenko N S.Technique of Gathering and Data Processing for an Estimation of Structure of Consumers of Services of a Tourist Complex of Region［J］.Practical Marketing，2009(11): 16－28.

［2］ Lewins A，Taylor C，Gibbs G R.What is Qualitative Data Analysis［EB/OL］.(2010－11－01)［2011－12－30］.http://onlineqda.hud.ac.uk/ Intro_QDA/what_is_qda.php.

［3］ Davidov A A.Qualitative Research:Develop-ment Prospects［EB/OL］. (2008－04－03)［2011－03－10］.http://www.isras.ru/index.php?page_id=922.

［4］ Kanygin G V.Konstruiruja Constructivism［J］.Sotsiol исслед，2006(11):19－28.

［5］ Gill Ereaut.What is Qualitative Research Sofeware［EB/OL］.(2008－05－10) ［2011－03－13］.http://www.qsrinternational.com/#tab_you.

［6］ Martyshenko N S，Vlasenko A A.Analys of Strategy of Development of a Tourist Complex of region［J］.Science Territory，2007，2(3): 175－182.

［7］ Martyshenko N S，Starkov A S.Analys of Structure of the Consumer Regional Tourist market［J］.Science Territory，2007，4(5):468－478.