Johann Mitlöhner

Software project effort estimation with voting rules

Social choice deals with aggregating the preferences of a number of voters into a collective preference. We will use this idea for software project effort estimation, substituting the voters by project attributes. Therefore, instead of supplying numeric values for various project attributes that are then used in regression or similar methods, a new project only needs to be placed into one ranking per attribute, necessitating only ordinal values. Using the resulting aggregate ranking the new project is again placed between other projects whose actual expended effort can be used to derive an estimation. In this paper we will present this method and extensions using weightings derived from genetic algorithms. We detail a validation based on several well-known data sets and show that estimation accuracy similar to classic methods can be achieved with considerably lower demands on input data.

Effort estimation for enterprise resource planning implementation projects using social choice-a comparative study

ERP implementation projects have received enormous attention in the last years, due to their importance for organisations, as well as the costs and risks involved. The estimation of effort and costs associated with new projects therefore is an important topic. Unfortunately, there is still a lack of models that can cope with the special characteristics of these projects. As the main focus lies in adapting and customising a complex system, and even changing the organisation, traditional models like COCOMO can not easily be applied. In this article, we will apply effort estimation based on social choice in this context. Social choice deals with aggregating the preferences of a number of voters into a collective preference, and we will apply this idea by substituting the voters by project attributes. Therefore, instead of supplying numeric values for various project attributes, a new project only needs to be placed into rankings per attribute, necessitating only ordinal values, and the resulting aggregate ranking can be used to derive an estimation. We will describe the estimation process using a data set of 39 projects, and compare the results to other approaches proposed in the literature.

A distance-based comparison of basic voting rules

In this paper we provide a comparison of different voting rules in a distance-based framework with the help of computer simulations. Taking into account the informational requirements to operate such voting rules and the outcomes of two well-known reference rules, we identify the Copeland rule as a good compromise between these two reference rules. It will be shown that the outcome of the Copeland rule is “close” to the outcomes of the reference rules, but it requires less informational input and has lower computational complexity.

Bounded rationality and adaptive agents in economic modeling

Traditional economic theory sees human economic decisions as rational choices of action to achieve maximum utility. However, economic reality shows that people often behave in ways different from that ideal: actions are not always determined rationally; often they are influenced by other factors, such as chance and tradition. In order to describe the behaviour of economic agents in the real world, we take irrationality and adaptation into account: we present work in progress on a market simulation that shows the effects of partly irrational and adaptive consumer buying decisions on the course of action of entrepreneurs.

Characteristics of Open Data CSV Files

This work analyzes an Open Data corpus containing 200K tabular resources with a total file size of 413 GB from a data consumer perspective. Our study shows that ~10% of the resources in Open Data portals are labelled as a tabular data of which only 50% can be considered CSV files. The study inspects the general shape of these tabular data, reports on column and row distribution, analyses the availability of (multiple) header rows and if a file contains multiple tables. In addition, we inspect and analyze the table column types, detect missing values and report about the distribution of the values.

Contrasting Rankings from Social Choice Aggregation Methods for Business Information System Selection in Multiple Case Studies

This article focuses on decision making for Information System (IS) selection, in particular in the context of business applications. The complexity of business IS is increasing continuously, covering a wide range of different aspects. These aspects can be adequately covered by multiple criteria in decision making methods, which have been developed for the last decades to support the evaluation, selection, and followup controlling of IS. The area of multiple attribute decision making (MADM), where problems are represented by several (conflicting) attributes or criteria, was described as the most well known branch of decision making [11]. These methods appeal to management due to their intuitive, simple and cost effective application. They are relatively transparent, allowing others to see the logic of the results and enabling the inclusion of the full range of intangible consequences in terms of considered attributes. Nevertheless, many difficulties in MADM exist that lead to application errors in business practice. Model extensions proposed to avoid these errors are in general coupled with increased complexity that in turn hinders their acceptance. In this article, we seek to increase the awareness and analyze the applicability of well researched selection rules originating from social choice theory in the context of business IS decisions. In particular, we seek to contrast rankings from various social choice aggregation methods and compare the outcomes with results from traditional utility ranking methods applied in practice considering two ex-post case studies.

Using Social Choice Rule Sets in Multiple Attribute Decision Making for Information System Selection

The evaluation of investments in information systems (IS) is usually based on conflicting criteria applied to the available alternatives, and the results are aggregated into a single ranking. The aggregation process is regularly complicated and biased through the usage of criteria weights. This article simply suggests avoiding the weighting process, and alternatively relies on a set of multiple social choice methods. This work investigates various methods of social choice voting rules for aggregation and the properties of the results they deliver in typical IS decisions. Results are compared with the outcome of traditional multiple attribute decision making, taking into account case study and simulation data. The results support our notion that weighting criteria in the context of complex IS investment appraisals does not provide a different or more comprehensive outcome than the less demanding social choice rule set applied.

Porting APL-programs via ASCII-transliteration

A scheme for translating APL characters to ASCII text is introduced, and an APL application is developed which reads and writes objects to and from files using the scheme. The system is available freely for several APL dialects; it can be used to port APL applications across different environments. The translation scheme used is proposed as a standard for the ASCII representation of APL characters, and feedback on this scheme is invited.

Semantic web and natural language in online discussion forums

Personal weg pages, blogging services and discussion forums have gained widespread acceptance among casual Internet users to communicate their views and preferences across diverse areas of topics. Unfortunately, since these contributions are written in free-form natural language it is difficult and error-prone to automatically extract much more than a classification of postings into subject categories; the statements contained in the individual sentences of the text remain largely inaccessible to automated analysis. The RDF Resource Description Framework provides sophisticated vocabulary for semantic markup, however, due its complexity it is not suitable for casual users. In this work an approach similar to the historical evolution of trade languages among different native speakers is suggested that aims at merging RDF and natural language into simple community-specific web pidgin languages interactively evolving based on the input of participating users.

Gathering Preference Data from Restricted Natural Language

The Resource Description Framework provides elaborate methods of specifying semantic information for automated aggregation and other types of processing; however, RDF is too complicated for most casual computer and internet users who prefer instead to communicate in natural language text form. Recently, much progress has been made in the area of extracting semantic data from unrestricted natural language, such as automatic document classification; however, much of the content of natural language texts still remains inaccessible for automated semantic processing. In this work a restricted natural language approach to the expression of preferences is proposed that sacrifices freedom of expression for accurateness of semantic translation. A limited number of grammatical constructs for the expression of preferences is described, and an application to the expression of cinema and movie related preferences is outlined. It is shown that with a moderate amount of restriction on natural language at least in the domain of preference specification a large amount of semantic information can be extracted while maintaining a level of user-friendliness suitable for more wide-spread application beyond the limited number of RDF experts.