Exploratory Data Analysis “EDA” and Automotive Semiconductors – The Impact of Data Science on Automotive Semiconductors

In recent years, more and more data science has been introduced in various fields. And it is no exception in the field of automotive semiconductors. Here, we have described exploratory data analysis (EDA), one of the key themes of data science here, and explained the relationship between automotive semiconductors and EDA.

Exploratory Data Analysis (EDA) and Data Science

What is Exploratory Data Analysis (EDA)?

Exploratory Data Analysis (EDA) refers to “a method or approach for exploring and understanding a data set as part of the data analysis process. In other words, it is a method or approach for organizing data and understanding that data.

This EDA is often discussed within the context of a field called “data science”. And in recent years, this data science approach and methodology has been introduced in a variety of fields.

Exploratory Data Analysis and Data Science

So what is data science? Data science is defined as “a combination of diverse techniques, methods, and processes for collecting and analyzing data and gaining insights to solve problems and support decision making.

In other words, it is a series of processes to find order in a group of disordered data, discover the laws of the data from that order, create a model, formulate a hypothesis, and then test and confirm it. The data science process is divided into two stages: the stage of organizing disordered data and formulating a hypothesis or model, and the stage of testing that hypothesis.

Of these, EDA refers to the process and methods used to derive a law from unordered data and formulate a hypothesis. In contrast, CDA (Confirmatory Data Analysis) refers to the process and methods used to confirm or verify a hypothesis after it has been formulated.

Specific examples of exploratory data analysis methods

There are a variety of EDA methods, but only an exhaustive overview is given here.

Data Summarization:
This method calculates basic statistics of data to understand trends and variability. Basic statistics include mean, mode, median, variance, etc. These are the basic statistical analysis methods learned in junior high school and high school. It is a simple concept, but it is very important because it is the basis for various advanced methods.

∙ Data Visualization:
Visualize data using visual techniques such as graphs, plots, histograms, etc. to visually identify patterns and outliers. This allows you to sensibly capture trends in your data.

Data Distribution:
Examines how the visualized data is distributed and identifies normal, uniform, log-normal, etc. distributions. For example, it is used in production processes to represent product data in terms of distribution to estimate defects between processes.

Outlier detection:
identifies outliers and outliers in a group of data and evaluates their impact on data analysis.

Correlation Analysis:
Examines correlations between variables in data to understand relationships between variables. In some cases, graphs are used for analysis. Typical methods include “multivariate analysis. Least Squares Approximation” and others also include this concept.

∙ Pattern discovery:
looks for patterns or trends in the data and identifies whether different factors affect the data. In recent years, machine learning (artificial intelligence) may be used.

・Group analysis:
This is a method of dividing data into different groups and comparing the characteristics of each group to separate them into different groups. Typical examples include “cluster analysis,” and in recent years, machine learning (artificial intelligence) is sometimes used to improve efficiency.

Handling Missing Values:
Check for the presence of missing values and consider strategies to supplement or properly handle missing values. Typical examples include “least squares approximation” and “principal component analysis. In recent years, machine learning (artificial intelligence) may also be used.

Data Science Expertise

In many cases, dedicated statistical processing software is available for these processes, so that calculations are not done manually using a “function calculator” as in the past, but it takes a certain amount of skill and expertise to master these software programs.

And those who understand and can actually utilize such data science concepts and methods are called “data scientists.

Data scientists were introduced as “the most attractive profession of the 21st century” by a business magazine, and many companies are currently in the market for data scientists.

Specific examples of exploratory data analysis and semiconductor manufacturing and development

Let’s take a look at how EDA is used in the field of semiconductor manufacturing and development.

Process Control and Quality Control:
EDA is used to analyze data collected during the manufacturing process to ensure the quality of semiconductor devices.

Design and Verification:
EDA can be used during the design phase of semiconductor devices to simulate circuit operation and performance for optimal design.

Reliability Assessment:
Reliability assessment of semiconductor devices helps to predict device life and analyze data to improve reliability.

・Optimization of device characteristics:
EDA is used to optimize device performance. For example, data analysis is performed to optimize characteristics such as energy efficiency, speed, and signal-to-noise ratio.

Failure Analysis:
After a semiconductor device is placed on the market, failure data is collected and analyzed and used to identify the cause of the problem.

Visualization of data:
Visualization of semiconductor device data facilitates anomaly detection and discovery of data patterns, contributing to quality improvement.

Exploratory Data Analysis and Automotive Semiconductor Device Applications

Let me briefly explain some specific examples of EDA’s involvement in automotive semiconductor device applications.

Because automobiles require advanced technology and safety, a data-driven approach (data-driven approach) is especially important for controlling automobiles using onboard semiconductor devices. For such a data-driven approach, the EDA approach is useful to grasp the trends of the data generated.

With the spread of artificial intelligence, it is also conceivable that EDA processing could be incorporated into automobile control systems, where EDA processing could be performed autonomously without human intervention.

Specific examples:
– Automated driving technology:
Automated cars require many sensors and control units using on-board semiconductors; EDA is used to analyze sensor data and develop control algorithms to support the evolution of automated driving technology.

-Reliability and durability of sensors and devices:
The inside of a car is very hot in the summer and very cold in the winter. It is also a harsh operating environment with lots of vibration and dust. Therefore, in-vehicle sensors and in-vehicle semiconductor devices are especially demanding in terms of reliability and durability. EDA is then used to evaluate the reliability of sensors and devices.


EDA is used to analyze and optimize power profiles in HVs and EVs to help extend battery life.

Self-diagnostics and fault diagnosis:
Automotive semiconductor devices must have self-diagnostic capabilities, and EDA helps analyze data for fault diagnosis, increasing the safety of in-vehicle systems.

Data Security:
In recent years, security has become more important as vehicles are connected to networks, and EDA can help identify security weaknesses and take countermeasures.

In other words, the concept of EDA has been introduced in various situations in the automotive field. That is how much EDA and data science technologies are in demand in the automotive field.

summary

The above is an overview of exploratory data analysis and its relationship to semiconductor development and applications, particularly in automotive semiconductors.

Exploratory data analysis is a major theme of data science, but its actual application requires familiarity with both the basic concepts of data science and knowledge of the target product or service. For this reason, there is an urgent social need to train “data scientists” who are experts in this field. This is why liberal arts colleges and universities have established faculties in this field one after another in recent years.

There is an urgent need for human resources in this field in the automotive sector as well. If you are looking for jobs related to automotive semiconductors, please contact Turnpoint Consulting. About 80% of the jobs Turnpoint Consulting handles are private jobs. Therefore, we are able to introduce you to a large number of jobs with favorable conditions that are not posted on general job sites. We support your job change with our overwhelming knowledge of the industry and thorough interview preparation. Let Turnpoint Consulting help you find the right job at the right automotive company.

Author: Fumitoshi Sato Supervisor: Turnpoint Consulting Media Team
We are a media team that provides useful information about changing jobs related to the automotive and mobility industry. For those who are thinking about changing jobs, we will provide information about the job market in the automotive and mobility industry and how to prepare for the selection process, and for corporate recruiters, we will provide information about the flow of human resources in this industry. Our goal is to help you by providing information about changing jobs and recruitment in the automotive and mobility industry.