Weiss, S.M. & Indurkhya, N.: Predictive Data Mining. A Practical Guide. Morgan Kaufmann, San Francisco:CA, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....patterns of the data rather than to describe it, the rules must be comprehensible to use that knowledge for taking decisions and, finally, the volumes of data are frequently large. Classical approaches for data mining are propositional (statistical methods, neural networks, decision trees, [34]) in the sense that the rules or knowledge that these systems are able to obtain can only refer to one row at a time. More over, these methods usually work with only one relation (table) However, most databases have more than one table and many interesting rules which are to be learned are ....

S.M. Weiss and N. Indurkhya. Predictive Data-Mining. A Practical Guide. Morgan Kaufmann Publishers, San Francisco, California, 1997.

....will demonstrate a degrading performance for every redundant feature selected, while training times increase exponentially. It is essential to reduce the number of inputs to a neural network and to select the optimum number of inputs which are able to clearly define the input output mapping[7]. One of the features that is always forcibly selected is the result of the contingency applied on every valid operating point in the data warehouse. This is because this feature is necessary for training and testing purposes of the UMLPSE machine learning tools. E. The features randomisation ....

....this module all necessary parameterisation and initialisation of the machine learning tools is executed automatically according to the selected features, as well as the training of the tools themselves. The machine learning toolkit used in UMLPSE is the DMSK (Data Miner Software Kit) toolkit. [7] Its high modularity enables the addition of more automatic learning tools with practically no programming cost. In figure 4 the functions of the last three modules described are depicted in UML terms. Figure 4. UMLPSE Automatic Tools Training Sequence Diagram G. The machine learning tools ....

Weiss, S., Indurkhya N., Predictive data mining, Morgan Kaufmann, 1998.

....and library correspondingly. Classification is a very useful operation in problems where predictions for new cases can be made by looking at the cases from past experience with known answers. The examples of such problems are fraud detection, marketing, healthcare outcomes, investment analysis [WI98, page 7] automatic article and image classification. Di#erent techniques from statistics, information retrieval and data mining are used for classification. They include Bayesian methods[DH73, Lew98, FGG 97] Bayesian belief networks[Pea91] decision trees [Qui93] neural networks[RHW96] ....

S. M. Weiss and N. Indurkhua. Predictive Data Mining. Morgan Kaufmann, 1998.

....components and evaluate their performance. The Abbreviation Identifier We use a binary decision tree to model the characteristics of abbreviations. The advantage of decision trees is that they are highly explainable one can readily understand the features that are affecting the analysis (Weiss Indurkhya, 1998). Each node of the tree provides insight into the characteristics of the data. Furthermore, decision trees are well suited for combining a wide variety of information. Features used as input to the tree can range from linguistic information such as part of speech category to the frequency of a ....

Weiss, S. & Indurkhya, N. (1998). Predictive Data Mining. San Francisco, Morgan Kauffman Publishers.

....of tables into small (or at least manageable) tables, or binning. These algorithms all determine the predicitive model in two steps: First, the data is scanned one to three times and reduced to a manageable size. Second, the model is determined from the reduced data (Weiss and Indurkhya [13]) We suggest calling algorithms of this type reduction algorithms. The size of the reduced data set needs to be independent of the data size but typically will have an influence on the approximation error or bias. Reduction algorithms with a fully scalable reduction step are fully scalable. Note ....

S.M. Weiss and N. Indurkhya, Predictive data mining, a practical guide, Morgan Kaufmann Publishers, 1998.

....nodes. The work of Rivest [Riv87] presents a new representation, decision lists, that generalizes decision trees. The advantage of this representation is modularity and consequently interpretability. Generating Decision Lists Several algorithms appear in literature for building decision lists [Riv87, CN89, Coh95, Dom96, WI98]. In this dissertation we restrict to those approaches that build decision lists from decision trees, as it is done in [Qui93] In any decision tree when a case reaches a leaf, the conditions that must be satisfied appear along the path from the root to the leaf. So, any tree can be easily ....

Sholom Weiss and Nitin Indurkhya. Predictive Data Mining, a practical Guide. Morgan Kaufmann Publishers, 1998.

....of features for coding our domain. We take the predictive data mining point of view, and focus on finding those features that have predictive power (for a general discussion on various definitions of relevance , see (Blum Langley 1997) A recent pragmatic survey of this area can be found in (Weiss Indurkhya 1998). Feature selection is often motivated by performance issues: in many cases the purpose of feature selection is elimination of irrelevant or redundant features without sacrificing prediction performance. From purely theoretical standpoint, having more features should always give us better ....

Weiss, S., and Indurkhya, N. 1998. Predictive Data Mining.

No context found.

Weiss, S.M. & Indurkhya, N.: Predictive Data Mining. A Practical Guide. Morgan Kaufmann, San Francisco:CA, 1998.

No context found.

Weiss, S.M.; Indurkhya, N. (1998): Predictive Data Mining. Morgan Kaufmann Publishers, Inc., San Francisco.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC