Utilizing machine learning to build a competitor pricing tool for the semiconductor industry. Case company: Silicon Labs

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Business | Master's thesis

Date

2024

Department

Tieto- ja palvelujohtamisen laitos

Major/Subject

Mcode

Degree programme

Business analytics

Language

en

Pages

75+22

Series

Abstract

With an extremely positive outlook, the semiconductor industry is expected to grow exponentially as its products power the entire global economy. As an industry that is highly competitive and price-sensitive, pricing has always been the primary focus for many industry players, to defend or gain new market shares and improve profit margins. This thesis aims to utilize machine learning to predict competitor cost, providing the case company with a critical business process and valuable insights to support strategic pricing activities. The thesis is conducted in collaboration with Silicon Labs, a global semiconductor company with almost 30 years of history. The paper includes a literature review and a data analysis section. More specifically, the former delves into the semiconductor industry and its cost structure, followed by the concept of pricing, its roles in business management, and different machine learning approaches and limitations. On the other hand, the data analysis section examines the case company’s data on product features and COGS in order to build a linear regression model which can be utilized to predict competitor costs. The dataset used for the analysis contains 352 products, with a total of 188 features. However, after cleaning and pre-processing, the resulting dataset is reduced to 79 features. Additionally, the analysis also includes a technique of feature selection (LASSO), which resulted in a model with lower testing errors, signifying a model with improved performance on unseen data. The resulting business process and data analysis provide the case company with a tool to predict competitor costs, translating to several benefits for decision-makers, such as improved profitability, cost management, guided decisions in pricing, market positioning, and negotiation, as well as enhanced risk management. This also echoed in the author’s recommendations for further development by the case company, which includes establishing a standardized product database of both internal and competitor products to enable comprehensive competition analysis, as well as enhancing the existing lookup dashboard with additional product filters and advanced analytics tools to improve accuracy and usability.

Description

Thesis advisor

Peura, Heikki

Keywords

pricing, semiconductor, machine learning, linear regression, predictive analytics

Other note

Citation