1. IOT in Transport: Data to Information & Information to Knowledge
Wael Elrifai
Sr. Director of Enterprise Solutions

2. Q: What is the difference between a school teacher and a steam locomotive?
A: The school teacher tells you to spit out your gum, while the locomotive says "Choo Choo Choo!" 

3. 5 Cs of IoT Connections Conversions Centralisation Cognition
What Is The IoT? 5
Connections Cs of
IoT
Conversions
Centralisation
Cognition
Regulatory such as Uber and AirBnB do not implement novel ideas, this is not the sort of innovation I’m talking about. A decent way I like to think about these things – if they’re not patent eligible, they’re probably not that innovative (prior art + prior art does not a patent make).
It’s not just re-implementing an existing platform on a “Big Data” platform.
It’s not creating a capability without using it.
It is being able to create a simplification of the world around us, what we call a “model”, that can make meaningful predictions about outcomes given an manageable set of inputs.
It isn’t usually transformative at the outset…
Continuous improvement

4. Ask the following question…
“What could I do if I had perfect, universal, and timely information?”
Then work backwards from there, what would be required to get that data (not information, but data) – don’t ignore outside sources, third parties, public data, etc.
Sensors, vidoe

5. Innovating in Transport
Business Challenges
Modernize and improve rail transportation reliability
Reduce maintenance costs
Use Case
XaaS / Usage based pricing
Anomaly Detection
Predictive Maintenance
Schedule Optimisation
Total Asset Optimisation
Big data science is going to allow us to predict equipment failures saving on scheduled replacement of non-faulty parts, reducing in-service downtime, and optimising operating schedules.
This meets a growing client demand to shift from CapEx to XaaS and outcome- based pricing.
Pentaho will eventually be deployed inside the train itself, in what we in the field call “edge analytics”.
The platform will continue to grow, supporting more trains and more train operating companies
By the numbers
3000 sensors per train
5Hz sensors (sensors produce data 5 times per second)
241 trains in the first 5 years
3.6 million data points per second, 570 trillion in the first five years.
Up to £20m savings per year

6. A (Very Rough) Outline of Data Science
Plain-Old
Statistics

7. Plain-Old-Statistics
Linear & Logistic Regression
Cause of A/C Failure
Simple component RUL
QC/QA
Statisticians must understand how the data was collected, statistical properties of the estimator (p-value, unbiased estimators), the underlying distribution of the population they are studying and the kinds of properties you would expect if you did the experiment many times. You need to know precisely what you are doing and come up with parameters that will provide the predictive power. Statistical modeling techniques are usually applied to low dimensional data sets.
Machine learning requires no prior assumptions about the underlying relationships between the variables. You just have to throw in all the data you have, and the algorithm processes the data and discovers patterns, using which you can make predictions on the new data set. Machine learning treats an algorithm like a black box, as long it works. It is generally applied to high dimensional data sets, the more data you have, the more accurate your prediction is.

8. A (Very Rough) Outline of Data Science
Unsupervised
Learning
Plain-Old
Statistics
Unsupervised learning is the machine learning task of inferring a function to describe hidden structure from unlabeled data. Since the examples given to the learner are unlabeled, there is no error or reward signal to evaluate a potential solution. This distinguishes unsupervised learning from supervised learning and reinforcement learning.
Supervised learning is the machine learning task of inferring a function from labeled training data. The training data consist of a set of training examples. In supervised learning, each example is a pair consisting of an input object (typically a vector) and a desired output value (also called the supervisory signal). A supervised learning algorithm analyzes the training data and produces an inferred function, which can be used for mapping new examples. An optimal scenario will allow for the algorithm to correctly determine the class labels for unseen instances. This requires the learning algorithm to generalize from the training data to unseen situations in a "reasonable" way (see inductive bias).
collect information about the environment is by interacting with it. We don't have input output pairs.

9. Unsupervised Learning
Typical Operation
Anomaly Detection
Automated Clustering of Devices
Unsupervised learning
No target attributes.
We want to explore the data to find some intrinsic structures.

10. A (Very Rough) Outline of Data Science
Unsupervised
Learning
Supervised
Learning
Plain-Old
Statistics
Unsupervised learning is the machine learning task of inferring a function to describe hidden structure from unlabeled data. Since the examples given to the learner are unlabeled, there is no error or reward signal to evaluate a potential solution. This distinguishes unsupervised learning from supervised learning and reinforcement learning.
Supervised learning is the machine learning task of inferring a function from labeled training data. The training data consist of a set of training examples. In supervised learning, each example is a pair consisting of an input object (typically a vector) and a desired output value (also called the supervisory signal). A supervised learning algorithm analyzes the training data and produces an inferred function, which can be used for mapping new examples. An optimal scenario will allow for the algorithm to correctly determine the class labels for unseen instances. This requires the learning algorithm to generalize from the training data to unseen situations in a "reasonable" way (see inductive bias).
collect information about the environment is by interacting with it. We don't have input output pairs.

11. Supervised Learning Complex system degradation Root cause analysis
Doors
HVAC
Bathrooms
Root cause analysis
Delay fault attribution

12. A (Very Rough) Outline of Data Science
Unsupervised
Learning
Supervised
Learning
Plain-Old
Statistics
Reinforcement
Learning
Unsupervised learning is the machine learning task of inferring a function to describe hidden structure from unlabeled data. Since the examples given to the learner are unlabeled, there is no error or reward signal to evaluate a potential solution. This distinguishes unsupervised learning from supervised learning and reinforcement learning.
Supervised learning is the machine learning task of inferring a function from labeled training data. The training data consist of a set of training examples. In supervised learning, each example is a pair consisting of an input object (typically a vector) and a desired output value (also called the supervisory signal). A supervised learning algorithm analyzes the training data and produces an inferred function, which can be used for mapping new examples. An optimal scenario will allow for the algorithm to correctly determine the class labels for unseen instances. This requires the learning algorithm to generalize from the training data to unseen situations in a "reasonable" way (see inductive bias).
collect information about the environment is by interacting with it. We don't have input output pairs.

13. Reinforcement Learning
Optimise complex systems
HVAC
Scheduling
Simulation & Design
Reinforcement learning is learning what to do i.e. how to map situations to actions--so as to maximize a numerical reward signal. Unlike most forms of machine learning, the learner is not told which actions to take. Instead, the learner must discover which actions yield the most reward by trying them. Actions may affect not only the immediate reward but also all subsequent rewards. Thus, the two characteristics: trial and error search and delayed reward are the two most important distinguishing features of reinforcement learning.
Reinforcement learning works in a cycle of sense-action-goals. Because reinforcement learning learns from immediate interaction with the environment, it is different from supervised learning (learning from examples provided by a knowledgeable external supervisor).  The interactive learning approach is beneficial in navigating in uncharted territory. Here, ‘uncharted territory’ refers to situations where we
We don’t know the examples to train on
We don’t know the right and wrong values for the examples
However, we do know an overall goal and
We can sense the environment and take steps to maximise both immediate and long term rewards.
One of the challenges that arise in reinforcement learning and not in other kinds of learning is the trade-off between exploration and exploitation. The agent must perform both exploration and exploitation simultaneously.  At the same time, the agent must consider the whole problem and operate in an environment of uncertainty.

14. Your Innovation Toolkit
Data Capture Systems
Communications Platform
Data Storage & Computation Platform
Machine Learning & Evolutionary Algorithms
GUI-based Abstraction Tools
Don’t be “fooled” by visualisations
humans are adept at pattern detection, even false patterns.
Viz where does it work maps? A great hammer is a terrible saw

15. Thank You