Internet of Things

Among the aspects of social analytics, motion detection and facial recognition particularly

stand out. Motion detection can be used to assess the public interest in fashion shows where it

observes input from a person like their mouth and eye position to calculate their interest and

focus for a subject. This can open new opportunities for producing tailored video clips.

Quick Challenge

List at least five algorithms that can help in social analytics.

Security

IoT and data analytics can join hands to provide video analytics for security. Video surveillance

is hampered by issues because sometimes the criminal masterminds are able to trick the

authorities through deception techniques. By using analytics, businesses are trying to look

for “anomalies” in the video footages. Whenever any abnormal data is collected by the IoT

devices, they instantly notify the authorities. Moreover, it is possible to increase or decrease

the level of observed patterns in such video analytics by modifying it according to security

management.

Use of Machine Learning in IoT with Real Life

Machine Learning is a subset of AI which makes computer applications more precise with their

predictions. What makes it unique is that it does not require explicit programming to run.

MLborrows elements from predictive analytics and data mining. Similar to them, it goes through

data for the identification of patterns and modifies itself. The example of Machine Learning can

be easily observed in daily internet use. Have you ever noted how you get relevant items for your

online shopping? Similarly, whenever you are on Netflix or another movie streamingwebsite,

you get recommendations according to your history. This filtering and personalization run

through recommendation engines, an application of ML.

ML algorithms fall into the category of supervised and unsupervised learning. The former

category needs an experienced data engineer or analyst to generate the input and the required

output. Data analysts think about the variables or features that require analysis by the model

for the process of calculating predictions. On the other hand, unsupervised learning is done

without the help of a professional. Clustering in machine learning or K-means algorithm is an

example of the unsupervised algorithm. It makes use of deep learning (iterative approach) to

process data and generates conclusions.

Let’s look over some real-life examples of ML.

Smart Home

The ML and IoT revolution will start from your homes! From smartphones, we have gone over

to smart homes. The dierence between a standard home and smart home is simple: the latter

has appliances and devices with internet. Alarm clocks, refrigerators, TVs, and dishwashers

everything would be connected to the internet. Add ML in the mix, and your entire home would

learn from your habits and adapt accordingly. For instance, an alarm clock can program itself to

continuously ring daily to wake you up for oce while it remains disabled on the weekends as

ML-based algorithms learn the concept of holidays.

Chapter 10 Data Analytics and Machine Learning for IoT 251

Internet_of_Things_CH10_pp249-270.indd 251 9/3/2019 10:15:56 AM

Power Plug

Weather

Sensor

Vibration

Sensor

Smart

Light

Door

Sensor

Smoke

Detector

Inlet Switch

Smart

Gateway

Motion

Sensor

Smart Ring

Smart Shirt

with Heart & Respiration

Sensors Incide

Smart

Finger

Smart

Glasses

Smart

Bracelet

Smart

Watch

Bluetooth

Key Tracker

Smart

Belt

Smart

Pants

Smart

Shoes

Smart

Socks

Wearables

Over the last few years, wearables are in-demand. With IT giants like Samsung and Google

investing in wearables, why are they so important? Wearables are devices with IoT-based

equipment like sensors. They gather data about users and process it to provide useful insights.

Theyare used for entertainment, fitness, and health purposes. With machine learning, wear-

ables are saving lives. While monitoring the heart rate and other essential bodily functions,

252 Internet of Things

Internet_of_Things_CH10_pp249-270.indd 252 9/3/2019 10:15:56 AM