Prevention of Human Errors for High Service Availability - Techgig Webinar

In this post, I am going to give you an overview of webinar Prevention of Human Errors for High Service Availability by Thripthy Antony, Process Manager, IT Operations at SAP via Techgig platform.

First of all, let us understand the types of errors:

• Mistakes may happen in planning to maintain a sequence
• Slips which means missing one or more steps while executing
• Lapses i.e. Mix up of things

Postmortem:

To avoid the errors happening in the future, there is a need to do Postmortem or WHATT (What happened at that time) analysis of an occurred error. It’s a meeting attended by stakeholders and team that captures the following:

• Timeline – Time at which error happened
• Assumptions and Expectations occurred during the incident
• Emotions – Feelings while error happened and at the time of rectifying them
• Preliminary analysis
• Action Items to be taken

There are two things that might go wrong when an outage is attributed to human error:

1. Hindsight Bias – Which is an overestimation of abilities to achieve predicted outcomes
2. Outcome Bias – Error in evaluating the quality of decision

How to Avoid Bias and prevent Errors:

There are two techniques namely

• Five Why’s
• Swiss Cheese Model

Five Why’s: It helps in avoiding Hindsight bias by asking a series of minimum 5 Why questions revealing the root cause of human error happened.

Swiss Cheese Model: It’s another approach that looks at a gap in the 4 levels in order namely Organization level (Structure), Supervision level (Planning), Precondition (Circumstances) and Unsafe acts (Condition of people, Experience). This model provides an in-depth analysis to find root causes

Defining Action Items:

Action Items can be defined in two ways:

1. Checklist – Defining steps to follow for execution or activities
2. Poka-Yoke – The Japanese technique which ensures that preconditions are met

Action Items should both root causes for failure and success as well

Always know that Human Error is not a Root Cause:

Some of the methods to avoid root causes of Human Error maybe

• Enhancement of Reporting culture
• Make sure that employees have the freedom to report errors
• Believe in Continuous Improvement of the system
• Make Robust systems by making activities error-proof

Some General Tips and Additional resources:

Some common resources and tips can be seen below:

• Jira is the best tool to track errors or capture them as well as defining reporting structure
• Don’t have too many metrics displayed on the dashboard which leads to confusion
• Avoid things that cause human errors in the system
• Always learn from mistakes and use various colors for various commands
• Attach action items on the dashboard
• To understand more on human errors, study the book The Field Guide to Human Error Investigations by Sidney Dekker

Introduction to AI and Data Science

In this post, I am going to give you an overview of webinar Introduction to Artificial Intelligence and Data Science by Bhanu K, Co-founder & Chief Data Scientist at nQuantum via Techgig platform.

History of Artificial Intelligence (AI):

Let us have some basic understanding of the history of AI:

• One of the founders of AI is John McCarthy who held first a first academic conference in AI in the year 1956
• Alan Turing proposed “Turing Test” can be considered as a measure of Machine Learning which determines machine ability to think like a human 
• From 1990, Machine Learning (ML) and AI became popular

Technology Terms: 

Before going further on AI and Data Science, let us understand some of the key terms in the industry:

• Data Science: It means working with data in a scientifical manner. This is not a technology and it’s a platform or field of study
• Machine Learning: ML is nothing but a combinative usage of Statistical algorithms and Programming in Java, Python or R. It is the subset of AI
• Deep Learning: Deep Learning deals with the advanced concepts of ML and can be thought of as a subset of ML. Examples include technologies like Computer Vision, Neural networks, etc.
• Artificial Intelligence: This means providing intelligence to the device to do a certain prescribed activity that can be repetitive or intellectual. AI is the superset of both ML and Deep Learning

The relation between AI, ML, and some other domains:

Below is the diagram showing the relation between AI, ML, and other domains:

Some abbreviations used in the figure:

• NLP – Natural Language Processing
• CNN – Convolution Neural Network
• RNN – Recurrent Neural Network

Google Neural Machine Translation while translating one language to another may use English as an intermediate language (To translate Telugu to Tamil, first it converts to English which then got converted to Tamil)

 Data Science and AI in the IT Industry:

First, to process the data, we need it. Generally, Data is generated by the human who gives input to the applications. This is the area where Application developers and Application support engineers come into the picture.

Once the data is entered, it must be stored somewhere before processing. Now a day's data size is increasing at a huge rate. To manage this, we need people like Database developers and Database admins. Processing this stored data helps in decision making for a variety of businesses.

Also, there will be some stored past or historical data known as Dataware house. These are used to generate data models or schemas. In this area, we need people like Informatica developers to manage them.

Analyzing data is not sufficient. It should be represented to gain an advantage from it. To visualize or represent data there are tools like Tableau or Power BI. We need some BI people for this. Overall, Data Science is a platform for AI, ML and Deep Learning.

Who is Data Scientist?

A Data Scientist is a person who can understand, process, and analyze below forms of data:

• Textual Information
• Audio Data
• Images
• Videos

Steps to Become a Data Scientist:

• First, go through the Probability, Statistical concepts which involves studying techniques involved in both Inferential Statistics and Descriptive Statistics
• Then learn the Python / R / Java and the Data Visualization tool like Tableau. But rather than Java or R, Python became popular and easy to script with great community support
• By learning Math concepts, Programming language, Data Visualization one can become a Data Analyst
• After the above steps, go through the ML libraries like NumPy, Pandas, and Seaborn. Then go through ML Algorithms and how to do ML in cloud platforms without coding
• At this stage, one becomes an ML engineer
• To become a Data Scientist or AI Engineer, one needs to master industry skills like NLP, Recommendation Engines, Neural networks using open source technologies like Keras, PyTorch, TensorFlow, etc. 

Why AI is hyped now?

There are several reasons for this change. Some are listed below:

• Data Analysis needs huge computing power. With the recent developments in Computer Industry, this cost of computing reduced
• Hence more data can be stored and processed
• Also, these days semi-structured and unstructured data like Log files, Click Stream, Tweets, etc 
• Internet Connected Devices became increasing at a rapid rate
• Huge sizes of data like Yottabytes (1024) and Bronobyte (1027) came into existence

Applications of AI:

Some of the industries and its examples where AI can play a huge role is given below:

• Travel Industry
• Predicting the delay of Flights
• Dynamic Pricing of tickets on a demand basis
• Marketing
• Cross-selling & Upselling
• Predicting the lifetime value of a customer
• Health Care
• Disease Prediction
• Effective Medication
• Sales
• Offering Discounts
• Demand Forecasting
• Social Media
• Digital Marketing
• Sentiment Analysis
• Automation
• Self-Driving Cars
• Pilotless Drones and Aircrafts
• Credit and Insurance
• Claims Prediction
• Fraud and Risk detection

Common ML Deliverables:

ML can provide the below-mentioned things

• Predict output based on Input values
• Classification of objects
• Recommendation of things (Ex: Amazon, Netflix)
• Detection of Pattern and Grouping
• Anomaly Detection
• Object Recognition
• Actionable Insights
• Segmentation and Forecasting

Areas of AI:

Some areas of AI are listed below:

• Natural Language Processing
• Deep Learning
• Simulation Modeling
• Machine Translation
• Virtual Private Assistant
• Social Network Analysis
• IoT

Some Tips and Other Information:

Here are some tips and information regarding AI and Data Science

• Always go through and learn the open-source and industry-leading technologies
• For Open Data Sets, one can refer to California Open Data Sets and Kaggle
• To know about Chatbot, go through this opensource tool RASA
• For developing Python ML programs, it's good to use the Spyder or Jupiter IDE
• Keras library is used for building Neural networks and Tensor flow can be used as a backend in case of application development
• To become a Big Data Scientist, one needs a focus on technologies like Hadoop, Spark, etc