Predicting the A-Grade fabric rate is an essential and important step during the whole fabric manufacturing process. The traditional prediction method that Esquel Group has been using relies on both professional experience and a time consuming empirical references procedure from past year data. The prediction performance of the traditional approach is good but requires professional manpower and fails to identify the linkage between the defects and fabric features. In our project, we propose a high dimensional classification-regression hybrid model for predicting the A-Grade rate. A user-friendly software was developed to automatically produce the prediction for different fabric features, and can be easily implemented by a layman in minutes. Further implementation shows that our model outperforms the traditional method in terms of prediction accuracy.
Our prediction is built on a hybrid statistical model, and the estimation of the model relies on very recent research results in high dimensional discriminant analysis and high dimensional regression. The original data is irregular in that (i) the actual A-Grade fabric rate, which is also the outcome variable in our model, is a 1-inflated variable in the interval [0,1]; (ii) the covariate is of very high dimension. By formulating the 1-inflated outcomes as a high dimensional classification problem, and the prediction of non-one rates via a high dimensional generalized regression with logit link, we established a hybrid model for prediction the A-Grade fabric rate. We use penalization methods to obtain both model selection and estimation in our model. With expertise in programming, our team has also developed a user-friendly software that has now been used by Esquel Group in their factories.
Our approach is model based and is dynamically updated, providing Esquel with real-time improvements to the algorithm. Our model has identified important features that would affect the A-Grade rate, providing the company with valuable information about quality production. Feedback from Esquel confirms that the PolyU model outperform traditional methods. The model marked a fabric manufacture efficiency improvement of 0.11%, translating to savings of around 120,000 yards of fabric and nearly HK$2.5M in production cost savings annually. Traditional industries rely heavily on manpower, and often do not efficiently utilize the information hidden within historical data. The Esquel Group would like to pursue ‘digitalisation’ to revolutionise their manufacturing technology, and this project is one of the first data analytic projects built following their new digitalisation policy. Esquel states that our software has “largely transformed (a) vital step” of the manufacturing process, and more broadly “shifts the paradigm and changes the mindset of our workforce” regarding the future digitalisation of the manufacturing process.
