[Case Study] Machine Learning Example at a Coating Material Manufacturer
Introducing case studies of machine learning using "Pipeline Pilot"【Case Studies】
This time, the goal was to grasp the trends of optimal experimental conditions by predicting characteristic values. <Background> From traditional to modern times, developing materials with better characteristic values is an important task across all fields and remains a significant challenge. Additionally, since the characteristic values of materials can be influenced by various factors, there is a demand for increased efficiency in material development. For instance, factors such as the materials used during development, reaction time, temperature, and hysteresis may also affect the results. While it is ideal to introduce experimental design methods for all these conditions and discover optimal conditions through actual experiments and measurements, this is not realistic from a cost and time perspective. Therefore, we conducted machine learning using a small dataset that recorded several experimental conditions (for example, explanatory variables X1, X2, X3, X4, ...) and their respective characteristic values (for example, objective variables Y1, Y2). In this case, we were able to easily perform machine learning to create a model and improve prediction accuracy. *For more details, please contact us.*
basic information
This is a powerful software that aggregates data such as text, images, molecular structures, and databases, enabling effective utilization of various data by leveraging machine learning and statistical analysis. It also features integration capabilities with Python and R, and can efficiently perform Bayesian optimization, which has recently gained attention. *For more details, please contact us.*
Price range
Delivery Time
Applications/Examples of results
We have a proven track record of implementation in a wide variety of scenes, including material manufacturers, manufacturing industries engaged in material development, and related manufacturers in semiconductors and surface treatment.
Detailed information
-
Here is the original data.
-
In "Pipeline Pilot," we achieve a flow that simultaneously performs data shaping and machine learning without coding. "Read data, perform machine learning, conduct validation, and output reports." It is possible to create this unified flow in an easy-to-understand manner. In this case, the machine learning model created by the workflow was determined to have room for improvement based on the validation results. Therefore, we improved the accuracy of the machine learning model.
-
Using "Pipeline Pilot" makes it possible to "easily perform machine learning." In this case study, we were able to create a machine learning model and improve its predictive accuracy. Additionally, one of the primary applications of the created learning model is to "grasp the trends of optimal experimental conditions by predicting characteristic values."
