AI-supported bridge tendering
| dc.contributor.author | Boman, Daniel | |
| dc.contributor.author | Wallin, Lisa | |
| dc.contributor.department | Chalmers tekniska högskola / Institutionen för teknikens ekonomi och organisation | sv |
| dc.contributor.examiner | Sunesson, Kaj | |
| dc.contributor.supervisor | Sunesson, Kaj | |
| dc.date.accessioned | 2020-06-22T09:16:48Z | |
| dc.date.available | 2020-06-22T09:16:48Z | |
| dc.date.issued | 2020 | sv |
| dc.date.submitted | 2019 | |
| dc.description.abstract | The tender process is an important part of the construction industry where most projects are procured using a competitive tender. A tender bid is costly to calculate which means that for each tendering without receiving a contract, the cost of it needs to be accounted for in following projects, which makes it increasingly difficult to place competitive tenders. Due to the national principle of public access, all tender documents procured from the Swedish transport administration, Trafikverket, are public which underlines the perception that the construction industry deals with large amounts of data. However, the industry is well known for its fragmented data practices. Hence, a large potential lies within utilizing the available data and finding ways to transform data into information. Today's tender process is mainly based on the intuition of experienced practitioners. Given the importance of competitive tenders and the amount of existing data, a construction company would benefit from a process using data to support tendering. This study aims to create a step-by-step process to transform data to information with the overall purpose to guide tendering. The first step is to investigate Trafikverket as a data source in terms of structure and quality of public data. Then, the data is processed into a neural network model predicting the number of tender bids for projects. By performing these two steps and combining them, a sketch of a process divided into steps for each required action is formed. By decomposing all actions and evaluating options and choices along the transformation, a process is constructed. Collecting data from Trafikverket is time consuming due to unstructured archiving. Unless Trafikverket standardizes a digital praxis for archiving documents, or if natural language processing techniques significantly increase in their performance, Trafikverket is not advisable as a source for large quantities of data. Moreover, the neural network model does not have enough predictive power given the data size and input variables used in this study. The performance of the neural network model is Root-Mean-Squared Error of 2.45 given only 41 observations. However, performance can likely be increased further by adding more data or by further identifying input variables that affect the number of tender bids. Finally, this study proposes and eight-step process which differs from previous processes since it accounts for the fact that much data in the construction industry are in documents rather than in a proper database. | sv |
| dc.identifier.coursecode | TEKX08 | sv |
| dc.identifier.uri | https://hdl.handle.net/20.500.12380/300938 | |
| dc.language.iso | eng | sv |
| dc.relation.ispartofseries | E2020_004 | sv |
| dc.setspec.uppsok | Technology | |
| dc.subject | Data Transfer | sv |
| dc.subject | Data-Information-Knowledge-Wisdom, | sv |
| dc.subject | Knowledge Discovery in Databases | sv |
| dc.subject | Neural Network Model | sv |
| dc.subject | Construction Industry | sv |
| dc.subject | Tendering | sv |
| dc.subject | Trafikverket | sv |
| dc.title | AI-supported bridge tendering | sv |
| dc.type.degree | Examensarbete för masterexamen | sv |
| dc.type.uppsok | H | |
| local.programme | Management and economics of innovation (MPMEI), MSc |