Mediaboss Marketing

An electrical lineman for Southern California Edison works on replacing a transformer as a whole ... [+] block is rewired. Long Beach, California. April 2014.

Analytics are typically viewed as an exercise in data, software and hardware. However, if the analytics are intended to influence decisions and actions, they are also an exercise in organizational change. Companies that dont view them as such are likely not to get much value from their analytics projects.

One organization that is pursuing analytics-based organizational change is Southern California Edison (SCE). One key focus of their activity is safety predictive analyticsunderstanding and predicting high risk work activities by the companys field employees that might lead to a life threatening and/or life altering incident causing injury or death. Safety issues, as you might expect, are fraught with organizational perilpolitics, lack of transparency, labor relations, and so forth. Even reporting a close call runs counter to typical organizational cultures. These organizational perils are a concern to SCE as well, but the company has created an approach to address them. SCE hasnt completely mastered safety predictive analytics and the requisite organizational changes, but its making great progress.

A Structure for Producing Analytical Change

Key to the success of the SCE approach is the structure of the analytical team that is addressing safety analytics. It is small, experienced, and integrated. Two of the key members of the team are Jeff Moore and Rosemary Perez, and they make a dynamic combination. Moore is a data scientist who works in the IT function; Perez works in Safety, Security, and Business Resiliency, and is a Predictive Analytics Advisor. In effect, Moore handles all the analytics and modeling activities on the project, and Perez, who has many years of experience in the field at SCE, leads the change management activities.

Steps to manage organizational change started at the beginning of the project and have persisted throughout it. One of the first objectives was to explain the model and variable insights to management. Outlining the range of possible outcomes allowed Perez and Moore to gain the support needed for a company wide deployment. Since Perez had relationships and trust in the districts, she could introduce the project concept to field management and staff without the concern about Why is Corporate here?. Perez noted that its important to be transparent when speaking with the teams. That trust has resulted in the district staffs willingness to listen and share their ideas on how best to deploy the model, to address missing variables and data, and to drive higher levels of adoption.

The team took all the time needed to get stakeholders engaged. Moore came into the project in the summer of 2018, and he was able to get a machine learning model up and running in a month or so, but presenting it, socializing it, and gaining buy-in for it took far longer. Moore and Perez met with executives of SCE in November and December of 2018. Within days of these meetings the safety model analytics project became a 2019 corporate goal for SCE. Safety was the companys number one priority, and it was willing to try innovative ideas to move it forward. For such a small team to have their work made into a corporate goal is unusual at SCE and elsewhere.

The Risk Model and its Findings

SCE now has an analytical risk-based framework, and risk scores for specific types of work activities and the context of the work. The model draws from a large data warehouse at SCE with work order data, structure characteristics, injury records, experience and training, and planning detail. All those factors were not previously linked, and there wasas is often the case with analyticsconsiderable data engineering necessary to pull together and relate the data.

The machine learning model scores activities that teams in the field perform, like setting a new pole or replacing an insulator. Each activity may be more or less dangerous depending on the time of year, day of the week, weather, crew size and composition, and so forth. Replacing a pole, for example, may be only a moderate risk task in itself, but when done on the side of a hill in the rain with a crane it becomes very high risk. Instead of generic safety messages to employees, SCE can now get much more specific by describing the risk of particular activities they perform on the job in a particular context.

As the model learns it will recommend specific approaches to reduce the risk of a job, like altering the crew mix or crew size, requiring additional management presence, using specific equipment or rigging to perform the work, or creating a longer power outage in order to do the job more slowly. The latter recommendation runs counter to the culture of not inconveniencing customers, but if the model specifically recommends it, then the teams will discuss the contributing factors as well as their years of experience to mitigate the risk before executing the work.

The project has led to several more general findings, which are of greatest interest to SCE executives. For example, management has long been interested in using data to understand changing safety risk profiles of the field teams over time as a result of increasing/decreasing workloads or as weather patterns change. While the predictive model considers more than 200 variables, the findings from the model have been summarized into the top fifteen distinct drivers of serious injury and fatality. Some shifting of variables is expected over time, but there has been great interest in better understanding the initial set of risk factors.

Deploying the Model and Needed Organizational Changes

Moore and Perez are in the early stages of deploying the model; theyve rolled it out to six of 35 districts thus far. Each district has a unique personality, and they dont want cookie-cutter answers on how to deploy in their district.

Moore, whose primary role was to create the model, said he has realized that safety analytics are not just about a model. I started out thinking it was about an algorithm, but I realized many other factors were involved in improving safety. Moore said that he gets some pressure to move on to analytics in other parts of the business, but in order to see your models come to life you have to go through this kind of process. And everyone at SCE believes the safety work is critical.

Perez, whose primary focus is change management, listed some of the organizational changes in deployment. There might be training issuesnot only on analytics, but also communication, leadership and ownership. There might be process concernshow we plan and communicate work. There may be technology concerns in using the system.

Perez also says the process of working with a district is critical. You cant just walk into a district and disrupt their work flow for no reason, she says. They want to know your purpose and your objective. We try to connect, show transparency, and build trust that we are here to help, that we are here to observe how they mitigate risk, to share our findings, and to see how the findings might be integrated into their work practices. We hope they will help us understand the complexity they face every day.

Both team members say they learn something every time they visit a district. Moore notes, You can only see the data you can see in the data warehousetime sheets, work orders, etc. But when you talk to the people who do the work, you learn a lot about how the data is created and applied. With each visit I understand the drivers better and the complexity of the work. I can also speak the language better with each district visit, and I understand the process and the equipment better as well.

With the findings from the model, Moore and Perez are beginning to work with another partner at SCEthe HR organization. It is responsible for defining work practices, training needs, standard operating procedures, and job aids. Each of these is potentially influenced by findings about safety risks, so the goal is to incorporate analytical results into the practices and procedures.

The team is already working to modify the model to incorporate new factorsone of which, not surprisingly given the situation in California, involves the risk of wildfires. Moore and Perez are also trying to create more integration of the risk scores with the work order system. They also plan to try to incorporate the risk model into other SCE business functions like Engineering, which might be able to lower the risk in the planning and construction of the electric grid. All in all, using data and analytics to improve safety is a time-consuming and multifaceted process, but what could be more important than reducing injury and fatality among SCE employees and work crews?

Read the original post:
Machine Learning And Organizational Change At Southern California Edison - Forbes

Machine learning typically is used to solve a host of diverse problems within an organization, extracting predictive knowledge from both structured and unstructured data and using them to deliver value. The technology has already made its way into different aspects of a business ranging from finding data patterns to detect anomalies and making recommendations. Machine learning helps organizations gain a competitive edge by processing a voluminous amount of data and applying complex computations.

With machine learning, companies can develop better applications according to their business requirements. This technology is mainly designed to make everything programmatic. Applications of machine learning have the potential to drive business outcomes that can extensively affect a companys bottom line. The rapid evolution of new techniques in recent years has further expanded the machine learning application to nearly boundless possibilities.

Industries relying on massive volumes of data are significantly leveraging machine learning techniques to process their data and to build models, strategize, and plan.

While implementing the effective application of machine learning enables businesses to grow, gain competitive advantage and prepare for the future, there are some key practical issues in machine learning and their business implications organizations must consider.

As machine learning significantly relies on data, the occurrence of noisy data can considerably impact any information prediction. Generally, data from a dataset carries extraneous and meaningless information which can significantly affect data analysis, clustering and association analysis. Having a lack of quality data can also restrain the capabilities of building ML models. In order to cope with quality data and noise, businesses need to apply better and effective machine learning strategies through data cleansing and overall processing of data.

There is no doubt that the development of machine learning has made it possible to learn directly from data rather than human knowledge with a strong emphasis on accuracy. However, the lack of the ability to explain or present data in understandable terms to a human, often called interpretability, is one of the biggest issues in machine learning. The introduction of possible biases in data has also led to ethical and legal issues with ML models. Theinterpretabilitylevels in the field of machine learning and algorithms may significantly vary. Some methods are human-compatible as they are highly interpretable, while some are too complex to apprehend, thus require ad-hoc methods to gain an interpretation.

In the context of supervised machine learning, an imbalanced dataset often involves two or more classes. There is an imbalance among labels in the training data in several real-world datasets. This imbalance in a dataset has the potential to affect the choice of learning, the process of selecting algorithms, model evaluation and verification. The models can even suffer large biases, and the learning will not be effective if the right techniques are not employed properly. ML algorithms can generate insufficient classifiers when faced with imbalanced datasets. When trying to resolve certain business challenges with imbalanced data sets, the classifiers produced by standard ML algorithms might not deliver precise outcomes.

Thus, to address imbalanced datasets requires strategies like enhancing classification algorithms or balancing classes in the training data before providing the data as input to machine learning algorithms.

Share This ArticleDo the sharing thingy

About AuthorMore info about author

Vivek Kumar is the President of Consumer Revenue at UpGrad, an online education platform providing industry oriented programs in collaboration with world-class institutes, some of which are MICA, IIIT Bangalore, BITS and various industry leaders which include MakeMyTrip, Ola, Flipkart to name a few. He has 19 years of experience in diversified industries like Consumer goods, Media, Technology Products and Ed-ucation Services. He has been leading businesses & multi-cultural teams with a consistent record of market-beating performance and building brand leadership. His previous engagement has been with Manipal Global Education services as Sr General Manager, Education Services (Digital Transformation Strategy & Global Expansion).

Read more:
Decoding Practical Problems and Business Implications of Machine Learning - Analytics Insight

The events of the past few months have taught us, among other things, how little control we have over our destiny. A major crisis, such as the Covid-19 pandemic, can come unexpectedly at any point of our lives and ruin everything we have worked to achieve. But as much as it may seem unfair, the unpredictability of future events and the constant change is the only thing that is certain.

Many things could have been done better. When it comes to the world of tech and AI many businesses will learn that it is worth investing in robust and bias-free machine learning solutions. On a larger scale, hopefully the world leaders will learn to take scientific data more seriously and with a greater sense of urgency. However, the reality is that no matter how much we invest in making our predictive modelling algorithms more accurate, change will always bring unexpected events our way.

The fact that change is the only thing we can be sure of is one of the key wisdoms of Vipassana meditation practice. Both good and bad things will keep on coming our way and we have to just observe and keep calm. Easier said than done? True, but meditation can help people find their path to equanimity. If you are not too familiar with meditation but have a good understanding of the world of AI and machine learning, there is an interesting connection between the two that can help you grasp the key principles of meditation practice.

Before going any further, it is necessary to clarify that meditation focuses on the brain and you cannot really compare a machine learning model to a human brain. To simplify, it would be a bit like comparing the first basic telephone from the 19th century with the latest iPhone 11 Pro. The original telephone was only capable of performing one task at a time whereas the iPhone is capable of multitasking and has complex functionalities which are not fully understood to most users. However, it is worth observing that the overarching process which describes many commonly used machine learning systems can also be used to describe the functionality of our brains.

A machine learning system consists of an input (i.e. data), algorithms that adapt and improve the more data you feed into them and an output which is a result of that process. Similarly with the brain there is an input in a form of sensory data (i.e. our senses, such as sight and sound) and neurons transmitting electrical signals in the brain that produce an outcome, such as your thoughts and actions.

Everything we experience throughout our lifetimes can be treated as input data and contributes to the shape and functionality of the brain. What is interesting in the case of a human brain is that some of the input data is processed consciously, however the majority happens sub-consciously without individual's awareness. This sub-conscious data processing in cognitive science is often referred to as priming and is a reason for another well-known concept in machine learning, namely bias.

People as well as algorithms are prone to making biased decisions. Some famous examples include: the familiarity bias[1] liking more what you already know, symmetry bias[2] perceiving symmetric faces as more attractive, other biases related to appearance such as perceiving wider male faces as less trustworthy[3], and many more incorrect or inaccurate inferences made based on first impressions. Everything we experience in life has an impact on our brain processing and therefore our decisions.

In machine learning, data scientists spend a lot of time and effort on data pre-processing and data mining to remove bias from the data. Similarly, Vipassana meditation practice focuses on peoples data input the five senses: sight, sound, smell, taste and touch. Throughout the meditation practice students are encouraged to sit still for hours at a time without any distraction and simply be aware of and observe the sensations of the body i.e., the data input. This is what is being fed into the brain at any given time, and should therefore require at least as much attention as the data fed into a machine learning system.

The overarching process is simple: you smell a flower -> feel a pleasant sensation in the brain -> you smile. The simplicity behind this input-output scenario (as well as many neuroscientific studies[4] which show that activity in the brain starts before people consciously realise what they are about to do) can help us understand and accept that the concept of conscious free will is an illusion[5].One of the key objectives of Vipassana practice is that the scientific laws that operate one's thoughts, feelings, judgements and sensations become clear. Life becomes characterised by increased awareness, non-delusion, self-control and peace[6].

The concept of free will and attaching too much importance to the idea of the self is a common source of unhappiness. Mr. Goenka, the Burmese-Indian teacher of Vipassana meditation points out that there is a tremendous amount of attachment towards this physical structure, this mental structure, by identifying oneself as I, I, I And the result is misery[7]. This is commonly seen in our society, people often attach their self-worth to imaginary physical or mental concepts such as their background, skin colour, religion, wealth or nationality. Too much focus on self-identity results in many social problems such as racism and identity politics.

Understanding the simplicity of the input-output scenario that describes our brain functionality can help us move beyond these made-up concepts that divide cultures and societies across the globe. Instead, we should perhaps take inspiration from a simple reinforcement learning system, reward the brain with positive experiences for ourselves and others and allow it to evolve in the direction of tolerance, understanding and compassion in order to find our path to equanimity.

[1] Newell, B. R., Lagnado, D. A., & Shanks, D. R. (2007). Straight choices: The psychology of decision making

[2] Little, A. C., Jones, B. C., Waitt, C., Tiddeman, B. P., Feinberg, D. R., Perrett, D. I., Apicella, C. L. & Marlow, F. W. (2008) Symmetry is related to sexual dimorphism in faces: data across culture and species

[3] Stirrat, M., & Perrett, D.I. (2010). Valid facial cues to cooperation and trust: Male facial width and trustworthiness.

[4] Haggard, P. (2008). Human volition: towards a neuroscience of will

[5] Wegner, D. M. (2002). The Illusion of Conscious Will. Bradford Books/MIT Press.

[6] Vipassana Meditation, As taught by S.N. Goenka in the tradition of Sayagyi U Ba Khin (https://www.dhamma.org/en/about/vipassana)

[7] Vipassana Meditation 10-day Course, S.N. Goenka

Read more:
Meditation And Machine Learning: A Guide To Acceptance And Equanimity - Forbes

Every now and then an opportunity comes along that you just cant pass up. One such opportunity that fell into my lap was when OReilly media reached out to me to see if I was interested in partnering on a collaborative book on the ethics that surround data science. For those who know me and follow my work, they have seen me calling for more focus on ethics for several years. Ive written blogs and papers on the topic, Ive given many conference presentations on the topic (including at Predictive Analytics World 2019!), and Ive had countless discussions with clients

To view this content OR subscribe for free

Already receive the Machine Learning Times emails?The Machine Learning Times now requires legacy email subscribers to upgrade their subscription - one time only - in order to attain a password-protected login and gain complete access.

Sign up for the Newsletter with your Choice of social media account:

Go here to see the original:
97 Things About Ethics Everyone In Data Science Should Know - Machine Learning Times - machine learning & data science news - The Predictive...

(Subscribe to our Today's Cache newsletter for a quick snapshot of top 5 tech stories. Click here to subscribe for free.)

Carnegie Mellon University (CMU) and Facebook AI Research (FAIR) have developed a semantic navigation system SemExp, to train robots to recognise objects, using machine learning.

Through SemExp, a robot is trained to differentiate between a kitchen table and an end table, while it is also able to understand where these objects are likely to be found. The process allows the navigation system to think strategically about how to search for something, said Devendra S. Chaplot, a Ph.D. student in CMU's Machine Learning Department, in a release.

Classical robotic navigation systems, explore a space by building a map showing obstacles. The robot eventually gets to where it needs to go, but the route can be circuitous. The system uses its semantic insights to determine the best places to look for a specific object, Chaplot added.

By making the system modular, the overall efficiency has gone up. The robots can now focus on learning the relationships between objects and room layouts. It also enables the robot to navigate its way from point A to point B, in the quickest possible manner.

Going forward a navigation technology like this could improve the interactions between people and robots. While a robot could bring an item in a particular place or it could find its way when directed, said a CMU release.

You have reached your limit for free articles this month.

To get full access, please subscribe.

Already have an account ? Sign in

Show Less Plan

Find mobile-friendly version of articles from the day's newspaper in one easy-to-read list.

Move smoothly between articles as our pages load instantly.

Enjoy reading as many articles as you wish without any limitations.

A one-stop-shop for seeing the latest updates, and managing your preferences.

A select list of articles that match your interests and tastes.

We brief you on the latest and most important developments, three times a day.

*Our Digital Subscription plans do not currently include the e-paper ,crossword, iPhone, iPad mobile applications and print. Our plans enhance your reading experience.

Originally posted here:
A navigation system powered by machine learning is training robots to recognise objects - The Hindu