Original article: http://www.havlena.net/en/business-analytics-intelligence/predictive-analytics-project-in-automotive-industry/
I had a chance to work on a predictive analytics project for a US car manufacturer. The goal of the project was to evaluate the feasibility to use Big Data analysis solutions for manufacturing to solve different operational needs. The objective was to determine a business case and identify a technical solution (vendor). Our task was to analyze production history data and predict car inspection failures from the production line. We obtained historical data on defects on the car, how the car moved along the assembly line and car specific information like engine type, model, color, transmission type, and so on. The data covered the whole manufacturing history for one year. We used IBM BigInsights and SPSS Modeler to make the predictions.
2. Project Charter
Evaluate the feasibility of using Big Data analytics solutions for
Manufacturing to solve the problem of Predictive Vehicle
Inspection:
● Analyzing vehicle production history to predict car inspection
failures from the production line.
● Production shifts, specific employee, and other factors
The two Big Data Analytics solutions to be evaluated:
● IBM BigInsights
● Datameer 2.1
3. Approach & Proposed Solution
● Recognized the problem as a classification problem
similar to credit scoring or fraud detection.
● Classification is the problem of identifying to which of a
set of categories a new observation belongs, on the basis
of a training set of data containing observations whose
category membership is known.
● Build a predictive model based on machine learning
classification (supervised learning) to identify whether a
vehicle can be classified as good (passes quality check
on 1st try) or bad (fails quality check on 1st try)
4. Proposed Solutions - Tools
● BigInsights + SPSS Modeler
○ Hadoop is used to store big data and execute data
processing jobs in an efficient and distributed
fashion. IBM provides BigInsights as a management
and operational interface to simplify working with
Hadoop without doing much coding.
○ SPSS Modeler is a data analytics workbench that
allows the user to build predictive models by
leveraging built in algorithms and functions without
the need for programming
5. Proposed Solutions - Tools
● Datameer
○ Like BigInsights, Datameer Analytics Solution presents a
web based spreadsheet interface on top of a Hadoop
cluster and provides analytics functions and
visualizations out of the box without the need for writing
code.
○ DAS also has a Smart Analytics suite. One of the tools
available in that suite is a decision tree model which is a
descriptive model that can identify important factors that
affect quality.
○ Datameer can also be extended to run predictive models
created in R, SAS, SPSS, etc.
6. IBM Solution Architecture
SPSS Modeler
Client (only
Windows)
SPSS Modeler
Server (multiplatform)
SPSS Analytic Server
● allows analysts to do predictive analytics over big
data
● data centric architecture ensures scalability and
performance
SPSS Analytic Catalyst
● automatically discovers statistically interesting
relationships in data
● close the analytic specialist gap
● good in early discovery dataset stage (helps to
focus on important parts)
● automate some parts of CRISP-DM
SPSS Analytic
Server
(multiplatform)
SPSS Analytic
Catalyst
Hadoop
(BigInsights)
7. Prediction in SPSS Modeler
425 predictors
85.4% accuracy
(on the training dataset)
10. c5.0 Algorithm
● C5.o is an algorithm used to generate a decision tree
which can be used for classification therefore it is often
referred to as a statistical classifier
● A C5.0 model works by splitting the sample based on the
field that provides the maximum information gain. Each
subsample defined by the first split is then split again,
usually based on a different field, and the process
repeats until the subsamples cannot be split any further.
Finally, the lowest-level splits are reexamined, and those
that do not contribute significantly to the value of the
model are removed or pruned.
11. c5.0 Algorithm
● C5.0 models are quite robust in the presence of
problems such as missing data and large numbers of
input fields.
● They usually do not require long training times to
create. Because of the algorithm’s recursive nature it can
benefit from parallel processing.
● C5.0 offers the boosting method to increase accuracy of
classification
12. Datameer Analysis
● As previously mentioned Datameer has some built in
advanced analytics tools but most of them are in the
descriptive analytics area. The sole predictive analytics
tool they have is a specialized recommendation engine.
● Datameer can be extended to include predictive models
generated in tools like R, SAS, SPSS, etc. These take the
form of functions in DAS similar to the concept of
functions in Excel.
○ The disadvantage of this approach is that the hard work
of building the model is done without the support of big
data
○ Another disadvantage is the lack of tight integration that
is present in the IBM solution however you do get the
freedom to use any tool
13. Project Challenges & Opportunities
● Data understanding and formatting
● Time constraints
● More interaction with people on the ground
● More predictor data (diverse dataset is a key!)
○ Plant environment (temperature, humidity,
pressure)
○ Specific employees
○ Supplier & parts data
○ Warranty data