Add Thesis

Adopting Machine Learning in Small Companies

Written by Y.-F. Chow, J. Kennerberg

Paper category

Bachelor Thesis


Computer Science




Thesis: Machine Learning for All (ML4A) Model From the survey in Section 4.4 above, several challenges that need to be considered when conducting this research are identified, as shown in Table 1. These challenges need to be addressed to make ML truly applicable to the current ecosystem of small companies. The difficulty of acquiring ML capabilities, and more importantly, the data science talent in the aML project is considered a major challenge. Since small companies often cannot afford to explicitly hire such talents, they need to provide solutions to this challenge. Too little data to get valuable results is also common in small companies, so it is another challenge to consider. The survey results show that small companies also have low general knowledge of MLP, which makes the description of the different MLP stages an important challenge. In order for this research to be useful to a wider audience, this challenge needs to be considered. Bridging the knowledge gap about what machine learning can and cannot do is also a challenge considered in this research. It is important to consider this in order to overcome the initial difficulties identified by ML novices in understanding practical solutions to a given problem. Simplify the guide for small companies to adopt machine learning. ML4A contains two sub-models, a machine learning usage model and a process model called agile machine learning (AML). Machine learning uses models to identify potential machine learning solutions for certain business problems. The usage model is created for those who are not familiar with the subject and have little knowledge of what ML can actually do. Although it does not fully cover all possible solutions, it provides general insights into what can be done with ML given a specific business problem. AML is designed for small companies to be able to develop their own ML solutions. AML is a process model that integrates MLP into agile SEP, designed to meet the needs of small companies. In order for AML to be truly applicable to small companies, several challenges identified in Table 1 need to be considered. The next section further describes the machine learning usage model. The latter part of the machine learning usage model is the use of machine learning in the agile software process. In this section, topics such as the classic agile software process, the machine learning process, and the integration of machine learning processes in agile are discussed. These topics are relevant to this study because they are the cornerstones of combining Agile SEP with MLP. Therefore, this section provides an overview of the subject. This section introduces the submodel AML. The last section introduces a case study using ML4A. During the case study, the sub-model AML was integrated into a specific agile SEP, Scrum. 5.1 Machine learning usage model This section is dedicated to solving one of the challenges identified in Table 1. The findings in the literature review point to knowledge gaps in what machine learning can and cannot do. In small companies that are new to ML and want to adopt it, guidance may be needed to determine which solutions are feasible for their identified business problems. Therefore, ML4A provides a model for the use of machine learning in an effort to close this gap. Machine learning uses models to serve as a guide for companies to identify their business problems and provide solutions that can have an impact on their business. Before companies adopt machine learning, it is important to determine the issues that machine learning can actually have an impact [8]. The purpose of using models in machine learning is to let people who are not familiar with machine learning have a deep understanding of what machine learning can actually do given certain business problems. First, when the goal is clear, the company should start the machine learning process. As shown in Figure 6, the first step is to identify the problem area. Going down the decision tree, the problem areas become more specific to further meet business needs. Appendix 3 provides solutions for different problem areas. The decision tree is based on the work of Hayes et al. [11], which identified the most common problems in small businesses in the United States. The model in the paper [6] was similarly studied in Sweden, so it is relevant to this research. The identified problem areas have been successfully solved by the different solutions in Appendix 3, proving that these solutions can actually improve the conditions of these problems, if they cannot be solved together. It should be noted that the solutions provided in Appendix 3 do not cover all possible solutions, but rather provide a deeper understanding of what can be done. The structure of Appendix 3 is a table with three columns. These columns contain the name of the solution, the ML task to which the solution belongs, and a brief description of the solution. The ML taskbar is included to give ML system developers a starting point when choosing algorithms. These tables are divided into different problem areas in the decision tree of the machine learning usage model. In order to better understand how the constellation works, some examples are provided below; for example, recruiting employees, the company has identified its recruitment of employees as a business problem. With the help of machine learning model decision tree, they reach the leaf node "recruit employees". Matching it with the corresponding problem area in Appendix 3, the company found a potential ML solution to the problem in the form of application and resume review. Read Less