داده کاوی متنی برای مدیریت دانش صنعتی و طبقه بندی متنی: رویکرد کسب و کار گرا

Textual databases are useful sources of information and knowledge and if these are well utilised then issues related to future project management and product or service quality improvement may be resolved. A large part of corporate information, approximately 80%, is available in textual data formats. Text Classification techniques are well known for managing on-line sources of digital documents. The identification of key issues discussed within textual data and their classification into two different classes could help decision makers or knowledge workers to manage their future activities better. This research is relevant for most text based documents and is demonstrated on Post Project Reviews (PPRs) which are valuable source of information and knowledge. The application of textual data mining techniques for discovering useful knowledge and classifying textual data into different classes is a relatively new area of research. The research work presented in this paper is focused on the use of hybrid applications of text mining or textual data mining techniques to classify textual data into two different classes. The research applies clustering techniques at the first stage and Apriori Association Rule Mining at the second stage. The Apriori Association Rule of Mining is applied to generate Multiple Key Term Phrasal Knowledge Sequences (MKTPKS) which are later used for classification. Additionally, studies were made to improve the classification accuracies of the classifiers i.e. C4.5, K-NN, Naïve Bayes and Support Vector Machines (SVMs). The classification accuracies were measured and the results compared with those of a single term based classification model. The methodology proposed could be used to analyse any free formatted textual data and in the current research it has been demonstrated on an industrial dataset consisting of Post Project Reviews (PPRs) collected from the construction industry. The data or information available in these reviews is codified in multiple different formats but in the current research scenario only free formatted text documents are examined. Experiments showed that the performance of classifiers improved through adopting the proposed methodology.

In the current digital based economy a large amount of information is available in the form of textual data which can often be used more easily if it is categorised or classified into some predefined classes (Miao, Duan, Zhang, & Jiao, 2009). In any business or industrial environment corporate information may be available in multiple different formats, about 80% of which is in text documents (Yu, Wang, & Lai, 2005). This information exists in the form of descriptive data formats which include service reports about repair information, manufacturing quality documentation and customer help desk notes (Kornfein & Goldfrab, 2007). It is also often in the form of concise text formats, containing many industry specific terms and abbreviations. Both technical and manual efforts are needed to handle these information sources, to unearth the patterns and discover useful knowledge hidden within these resources (Kornfein & Goldfrab, 2007). Transformation of these useful sources of information into usable formats will help to improve future product or service quality and provide solutions to project management issues. Decision makers or knowledge workers may therefore be assisted and business decisions improved through the discovery of useful knowledge patterns. Identified knowledge can also be transferred from one project to another. This will ultimately help to cut the overhead costs of product or service quality improvement and project management. Therefore the purpose of these studies is to try to improve any business context where useful knowledge of previous experience can be discovered in reports or other documents. For example, if customers’ needs can be identified and classified then better future decisions can be made resulting in improved levels of customer satisfaction. The overall process of knowledge discovery in databases (KDD) is the identification of valid, novel, potentially useful and ultimately understandable patterns in data (Fayyad, Piatetsky-Shapiro, & Smyth, 1996). The term knowledge discovery from textual databases (KDT) is a little different to the general form of KDD and can be defined as discovering useful information and knowledge from textual databases through the application of data mining techniques (Han and Kamber, 2000 and Karanikas and Theodoulidis, 2002). However it shares the common methods of collecting information as raw data and processing it through the application of data mining techniques. Indeed, a three step process of data collection, pre-processing and applications of text mining techniques (Karanikas & Theodoulidis, 2002) is required. Text classification is an important approach to handling textual data or information in the overall process of knowledge discovery from textual databases. It has been a most promising area of research since the inception of the digital text based economy (Ikonomakis, Kotsiantis, & Tampakas, 2005). It is mainly used to classify text documents into predefined categories or classes based upon content and labelled training samples (Jinshu, Bofeng, & Xin, 2006). Text mining techniques have been widely used in various application fields like e-mail filtering, document management, customer needs identification, etc. It can therefore be concluded that the use of this technology can help to access information and manage it for better use in future applications. Applications of data mining techniques have long been seen to improve predictive and classification methods and have widely been used in different subject areas ranging from finance to health sciences. Quite a few applications of these techniques have been reported in manufacturing or construction industry environments. There may however be problems of non-availability of data due to some confidentiality, proprietary and sensitivity issues (Wang, 2007). This leads to the exploitation of data mining techniques to handle textual databases being less frequently reported in the literature. The research work reported in this paper proposes a new hybridised method of handling textual data formats and classifying the text documents into two different classes. The effectiveness of the proposed methodology is demonstrated with the help of a case study taken from a real industrial context. The new approach adopted within this research will help to uncover useful information in terms of Multiple Key Term Phrasal Knowledge Sequences (MKTPKS) which can then be used for the classification of Post Project Reviews (PPRs) into good or bad information-containing documents. Focus has been put on the application of different classifiers such as Decision Trees, Naïve Bayesian learner, K-NN classifiers and Support Vector Machines (SVMs) to test the usefulness of the proposed model. The results obtained are also compared with simple bag of words (BoW) representation models and the F-measure is used as the quantitative metric for measuring the effectiveness of the model. The remainder of this paper is organised as follows: Section 2 provides the background for Text Classification methods and related work reported in the literature for industrial knowledge management solutions. Sections 3 discusses the proposed methodology and architecture, and different methods incorporated within this methodology. Section 4 discusses an implementation of the proposed methodology, based on real industrial data in the form of PPRs, and its classification results. Conclusions and future work are discussed in Section 5.

The research work presented in this paper is focused on classification of textual data into two different classes of defining good and bad information-containing documents. A novel integration of textual data mining techniques is made to improve the classification accuracies of the classifier. In terms of classifying documents into their respective classes using a decision tree (J48 or C4.5 algorithm) the accuracy of the classifier is reduced using the proposed methodology while in the other classifiers there is significant improvement in the classification accuracies measured using the F-measure. Further research is needed into improvement strategies for the proposed methodology which could be in terms of introducing more refined methods for matrix representation models. The reason behind the loss of accuracy in the case of C4.5 classifier may lie in the use of entropy based information selection criteria whereas other classifiers uses simple distance, probabilistic and kernel based distance measures to find the similarities between documents.