Techbyte-2019 | Insight

Machine Learning:Advantages And Applications

Faculty Mentor:
Dr.Deepshikha Aggarwal

Student Name:
Anjali Singhal(MCA – II)
Kritika sahni(MCA-II)

1.INTRODUCTION

Machine Learning, a technology that allows systems to learn directly from examples, data, and experience. It is the core sub area of Artificial Intelligence (AI). It is the learning and creation of algorithms that can learn by making predictions on data sets. It has become backbone of information technology in recent years. Machine learning algorithms creates a mathematical formula that is based on the data set for predictions. Making computers to work from experience like humans without explicitly performing direct programming. Basically, it’s an algorithm or model that learns patterns in big data and then predicts similar patterns in new data. Machine Learning is being implemented in multiple fields and businesses, and it is producing benefits. This hi-tech concept has ability to learn and improve customer experiences by analysing customer preferences and bring in productive outputs and results. This technique is extremely useful in helping humans perform complex tasks, such as predicting diseases, predicting stock market evolution, self-driving cars, Amazon, Face book, cyber fraud detection and many moreapplications to prove its worth.

2.EVOLUTION OF MACHINE LEARNING

The Machine Learning model was developed in 1949 by Donald Hebb in a book named THE ORGANIZATION OF BEHAVIOR. Arthur Samuel, from the field of computer gaming and artificial intelligence, originated the term "Machine Learning" in 1959 at IBM. Machine learning has evolved from pattern recognition and computational theory in artificial intelligence. It uses the in-depth study of algorithms and their construction for learning and making predictions or decisions. Machine learning often includes computational statistics and mathematical optimization which provide theory and methods for efficient decision making.

3.MACHINE LEARNING V/S TRADITIONAL PROGRAMMING

3.1 TRADITIONAL PROGRAMMING

Traditional computer programming involves multiple number of times user can use the software to perform the same task and it will always launch the restored position every time. Traditional Programming refers to a program that is manually written which takes input data and runs the program to produce the output.

3.2 MACHINE LEARNING

Machine Learning is called as augmented analytics where the input data and the output can find by an algorithm to create a computer program that can be used to figure the future outcomes. It helps to learn the software to gain the ability to learn from the previous observations and predict about the future behaviour and create possibilities of what can be done in new scenarios Machine learning is an automated process, the algorithms are automatically formulating all the rules from the data, which is a powerful concept. For example, if you feed in customer demographics or statistics, and the transactions take as input and the output if they moved in past or not and the algorithm will formulate the program which predict that someone would move or not.

4.WORKING OF ML

Machine learning is made up of three parts:

The computational algorithm at the core of making assurances.
Variables and features that contributes to the final outcome.
Basic knowledge for which the answer is known that build-up the system to learn.

Firstly, the model is provided with parameter data for which the answer is known. The algorithm is then run, and adjustments are made until the algorithm’s output admits with the known answer. At this point, increasing amounts of data are input to help the system learn and process higher computational decisions.

5.TYPES OF MACHINE LEARNING

There are two major types of machine learning: -

Supervised Learning: - “The output for the given input is already known before”. The machine must be able to map or assign the given input to the output. There is a need to supervise in these learning. A supervised algorithm gets output from labelled training data that helps you to predict outcomes for unforeseen data. It will also predict the data from the previous output. It will help you to optimize the performance by using their experience. It will help you to solve the real-world computation problem.
How Supervised Learning works: - For example, you want to create a machine that help you to figure how long it will take you to drive workplace from your home. Here you start by creating set of data like: -
eather conditions
Time of the day
No of holiday
Unsupervised Learning: - “The output for the given input is unknown”. There is no need to supervised Instead, you need to allow the model to work on its own to discover information. It mainly deals with the unlabelled data. In this type of learning this will allow you to solve complex problems as compared to supervised learning. It finds all type of unknown pattern in the given data. It can have helped you to find more characteristicwhich can be used for categorization.

6.IMPORTANCE

It allows people to do tasks more quickly and effectively.
It helps evaluating large chunks of data, easing the tasks of data scientists in an automated process.
It allows the machine to modify/change the tasks otherwise has to be performed by a human, example changing a password or checking an account balance.
It makes the machine to apply mathematical calculations automatically.
It supports data extraction and interpretation of the acquired data sets.

7.APPLICATIONS

Virtual Personal Assistant: -

Social Media Services: -

Product Recommendations: - When you shopped for a product online few days ago and then you keep recommendations on the web pages as shopping suggestions.
Predictions While Commuting: - We all are using GPS navigation services. While we are using GPS, our current locations are to be saved at a central server for managing traffic.
Robot Movement: -Self-supervised learning is used.
Detecting credit card illegal dishonesty:

•Customers were charged earlier for the items they did not purchase.

•Techniques -- Data mining and Bagging ensemble classifier.

And for many other almost the same computer program which require study of historic behaviour and possible future predictions machine learning is widely used.

8.MACHINE LEARNING METHODS

Regression: -Regression is the method of that’s comes in the category of Supervised Learning. They help you to predict a particular numerical based on the set of data. The simplest method is the linear regression and use the equation of line (y = m * x + b).
Classification: - The other method is classification that predict and explain the class value. The algorithm is the logistic regression it estimates the probability of an event based on the inputs.
Clustering: - The clustering method is got into the category of unsupervised machine learning. Its goal is to cluster the similar characteristics. we can only use visualizations to inspect the quality of the solution.
Ensemble: - This method combines several supervised machine learning to get the higher quality of data.
Neural Networks and Deep Learning: - In the contrast of linear and logistics regressions the purpose of the neural networks is to capture the non-linear patterns in data by adding parameters to the model.

9.ADVANTAGES AND DISADVANTAGES

ADVANTAGES

Easily identifies the Trends and Patterns: -Machine learning can track the large amount of data and finds the trends and the patterns that would not be done by humans.
Continuous Improvement: - They helps you to improving in accuracy and the efficiency that helps you to make faster and better decisions.
Handling multi-dimensional and multi-variety data: - Machine learning can handle multiple tasks and lager data sets easily in non-suitable and conditionals environments.
Social advertisements: - Social networking sites like Facebook, Twitter, Hike can promote much needed user history.
Allows real-world implementations: - Health, Banking, Financial sectors contributes a major amount of data to be further used.

DISADVANTAGES

Data Acquisition: - Machine learning needs huge amount of data and that should be of good quality. There can also be times where they must wait for new data to be generated.
Time and Resources: - Machine learning needs enough time to learn the algorithms and develop enough to complete their purpose with a considerable amount of accuracy. It requires massive resources to perform function.
Interpretation of Results: - Another challenge of machine learning is to get to know performance of machine learning algorithms.
Lack of variability: - It includes variance error and bias variance trade-off. It results in biased predictions.
Limited scope of success: - Most of the times machine learning will fail, thus it requires some understanding of the problem in order to apply the right machine learning algorithm.

10.CHALLENGES AND LIMITATIONS

Security, Privacy and Data Integrity
Dealing with Non-static, Unbalanced and Cost-sensitive Data.
Black box answers
Data set prices
Non ethical relativity
Less people with technical ability
Non self-explanatory machines
Requires larger space for data set storage
Time constraints
Limited predictions sometimes
Complicated when large data as requires more operations
Verification issues.

11.FUTURE SCOPE

Machine Learning proofs to be advantageous to any company including top MNC or a start-up or government offices are making things currently working properly and being done manually will be accomplished by machines. In the upcoming days, there will be much more demand of employees having the in-depth knowledge of machine learning in reputed companies like Google, Facebook, Twitter, Quora. With the huge population growing across the world, machine learning has wider scope of making peoples life easier by helping them out and proving them more efficient machines and robotic vision andpre-predict the upcoming future crisis which can adversely affect the world. Though humans have already been replaced by the machines in many household chores, industries, army and many more and it will be tremendously increasing in later years. Machine Learning is a technology which will stay with us in our future and so will be the future of Machine Learning.

12.REFERENCES

[1]http://papers.nips.cc/paper/3150-map-reduce-for-machine-learning-on-multicore.pdf
[2]https://www.upgrad.com/blog/scope-of-machine-learning/
[3]https://medium.com/deep-math-machine-learning-ai/different-types-of-machine-learning-and-their-types-34760b9128a2
[4]https://data-flair.training/blogs/advantages-and-disadvantages-of-machine-learning