Industrial discrete manufacturing processes meet the needs of business rates
Reserve early knowledge
Machine Learning Big Three: Numpy Matplotlib Pandas
tabular data mining algorithms: supervised, unsupervised
machine learning magic -Sklearn: application of machine learning algorithms Sklearn
Background - the revolutionary impact of Intelligent Manufacturing bring (Industry 4.0)
Business scene analysis
In the high-end manufacturing, with the deepening of the digital transition progresses, more and more data can be used to analyze and learn,
so as to realize significant manufacturing process control and decision-making aspects of intelligence, such as production and quality management.
Data-driven approach from the point of view, the production of quality management is usually required to complete quality factors affecting mining and predictive quality, quality control and other aspects of optimization,
this season will focus on the question, based on the relevant parameters and historical data on production potential with the first link analysis, complete quality-related factors and quality of the final confirmation in line with the predicted rate.
In actual production, the results of which will be an important part of the basis for the subsequent optimization of control
Topics
Job title race
Since the actual production, at the same set of process parameter setting will produce more artifacts quality results, we define its quality standards compliance rate for each set of process parameters, namely the set of quality parameters of the production process a workpiece excellent examination results are in line with the ratio of four types of indicators of good, acceptable and unacceptable. Compared quality prediction results of the respective workpiece, the quality standards coincidence rate prediction would be more meaningful.
This question asked participants to match the standard of quality for a given process parameter combinations produced workpieces compliance rate prediction.
Data Introduction DATA BACKGROUD
In this task, the workpiece to a typical production process as an example, we will provide a series of process parameters runners, and the quality of the produced data of the workpiece at the respective process parameters. The data comes from real data collected by a factory, we have done desensitization treatment.
(1) training data will provide:
A: process parameters (such as device processing parameters)
B: quality data of the workpiece
C: Quality Indicator workpiece conforms
(2) the test data is provided:
A: process parameters (such as device processing parameters)
Description DATA DESCRIPTION Data
(1) Preliminary training data set file name is first_round_training_data.csv, csv format, which includes fields 21, line 6000 containing A, B, C three types of data,
(2) Preliminary test data set file name is first_round_testing_data.csv, It contains 11 fields, the data only classes A,
Submission Requirements REQUIREMENTS ON WORKS
(1) Results of required files submitted from a csv format, which contains five fields, requires participants to provide test results data set groups all process parameters,
(2) Results of preliminary phase commit csv submit results: for algorithm contest, participants to csv file format, submit the results to large data model contest platform, the platform for online scores, real-time rankings.
Project ideas
1. Starting from the data
2.EDA found amazing point
3. Data preprocessing
4. Selection Algorithm
5. Deployment Model
Technical solutions - initially worked
Data Table DataMining Classification
Classification Algorithm Neural Network Ensemble Xgb Lgb Catboost
