DEV Community: Roksolana Kryshtanovych

Predictive Maintenance and Machine Learning

Roksolana Kryshtanovych — Thu, 10 Sep 2020 12:28:34 +0000

On the huge promise of predictive maintenance.

In industry, machinery downtime can lead to penalties, degrade a company’s core operations, and cause dire reputational damages.

It’s essential for both small enterprises and the Walmarts of the world to have a well-rounded, well-tested maintenance strategy in place to reduce the likelihood of sudden outages or breakdowns happening and mitigate all associated risks.

A proper approach to maintenance can help enhance, quickly, the overall reliability and performance of a company, while also substantially reducing the operational costs it incurs.

In 2018, Amazon had an outage that lasted for about 60 minutes and lost nearly $100m in sales as a result.

The recent proliferation of connected technologies and predictive machine learning algorithms has had a profound effect on how companies conduct their equipment management; these recent advancements have enabled firms to ditch, almost completely, the practices of traditional reactive (RM) and preventive (PM) modes of maintenance and start embracing data-driven predictive maintenance (PdM) instead.

In this article, we’ll explain accessibly the key differences between RM, PM, and PdM and discuss some of the AI models commonly used for condition-based machinery monitoring.

The post Predictive Maintenance: How Machine Learning Models Help Reduce Prevent/Repair Costs and Break Limitations of Traditional Maintenance Approaches appeared first on IT Consulting Company Perfectial.

Supply Chain Risk Management with Data Analytics

Roksolana Kryshtanovych — Tue, 25 Aug 2020 14:54:48 +0000

Everyone is talking about big data, but few are trying to benefit from it, notably in the supply chain risk management. According to statistics, 97% of supply chain analysts believe that big data can be extremely useful for the supply chain, but only 17% are actually using it. A supply chain is a great source of data, produced by customers, the business itself, and its operations. By analyzing and capitalizing on this data, businesses open for themselves the door to endless possibilities and obtain a considerable competitive advantage. So, let’s take a closer look at how the supply chain can benefit from big data analytics in risk management and how everything works.

Big Data Analytics for Supply Chain Risk Management: Why It’s Essential for Business

The incredible amount of data produced by the supply chain can be transformed into valuable insights to help identify both issues and opportunities to transform the company’s business strategy from reactive to proactive. Supply chain analytics use data and quantitative methods for enhanced decision making. It becomes possible with the evolution of datasets for analytics from the conventional, in many cases unstructured, data stored on both Enterprise Resource Planning and Supply Chain Management Systems.

These insights become especially important for supply chain risk management in the age of increased interconnectivity. The new risks, like cyber-threats, arise along with the traditional ones, making the supply chain more vulnerable than ever. Big data can help considerably in detecting and preventing these hazards. Moreover, the processing of supply chain data can contribute to improving customer service – it can help better preserve products during transportation and prevent shipment delays due to unforeseen circumstances.

The post Supply Chain Risk Management with Data Analytics appeared first on IT Consulting Company Perfectial.

PCA in Finance

Roksolana Kryshtanovych — Mon, 20 Jul 2020 13:24:15 +0000

Exploring one of the many applications of PCA in finance – stock price prediction.

In a recent post, we covered one of the most widely used data clustering algorithms K-means.

Here, we’ll shed some light on a bedrock technique for dimensionality reduction (applied extensively in probability and statistics) – principal component analysis (PCA).

What’s curious about PCA is that it’s not a dimensionality reduction method per se, despite mostly being advertised as such. Actually, it transforms data in a certain way to make it amenable to reduction later.

The algorithm finds new views (axes) for datasets in which it’s easy to separate out data instances and consequently cluster them.

By default, it creates as many axes as there are dimensions to a given dataset but it ranks these directions, called principal components (PCs), in the order of importance: the first PC always captures the most amount of data variance possible, the second one shows the second-largest amount of variance, and so forth.

After the data has been projected into this new subspace, we might drop out some of the axes and thus reduce dimensionality without losing much important information.

Let’s dig a bit deeper into this.

The post PCA in Finance appeared first on Bespoke Software Development Company Perfectial.

How Businesses Can Innovate in Retail with Latest Technology

Roksolana Kryshtanovych — Wed, 17 Jun 2020 10:53:42 +0000

With the ever-evolving consumer’s needs and requirements and the rapidly growing competition, it can be difficult for a retail business to keep up. New technologies are becoming a savior for this sector by granting the companies a range of exciting possibilities: from tools to create an unforgettable experience for customers both online and in-store to the chance of bringing innovation to the product creation. The retail software company can offer your business the true re-invention by conducting an in-store modernization or even complete digital transformation. So, let’s get to it and discover the latest technologies that innovate retail.

Retail Entering the New Age

The growing sales in retail and higher income in this sphere, make the business owners strive for taking their companies to another level by introducing the latest technologies to them. The social changes are not only driven by new technologies but, in many cases, they determine the innovations these technologies achieve and form consumers’ preferences. Digital technologies can disrupt the retail sector and help it to enter the new age by re-inventing the supply chains and introducing new business models.

One of the technologies that are a big deal for the retail digitalization is 4G (and pretty soon 5G). They give customers permanent connectivity and thus, grant more possibilities for retailers. The evolvement of online payments and the introduction of cloud computing are moving the retail sector even further. With the majority of people having easy access to software and data technology, it’s becoming possible for retailers to gather data on their customers to analyze and introduce improvements or make their experience more personalized. Another technological breakthrough in the retail business software is 3D printing which is predicted to be a huge innovation driver in the nearest future.

The post How Businesses Can Innovate in Retail with Latest Technology appeared first on IT Consulting Company Perfectial.

Intelligent Machine Learning Tools with Self-Organizing Maps and Other Unsupervised Methods

Roksolana Kryshtanovych — Fri, 05 Jun 2020 13:37:53 +0000

Most tech-oriented blogs have been largely one-sided in their coverage of machine learning, posting articles almost exclusively skewed towards supervised methods.

That’s not exactly unwarranted since supervised models, and particularly neural nets, have been deployed to great success both in business and academia (CNNs, for example, had been integrated into automated emergency braking systems, which helped reduce collisions by 40%), but it might’ve created the impression that there aren’t other ML subsets worth paying attention to, which is not the case.

It’s important to understand that all the algorithms (models, machines – whatever you want to call them) in supervised learning essentially do one thing – function approximation. They learn to map inputs to outputs after being fed tons of labeled data (e.g. being shown repeatedly what the desired outcome should be). They typically work with high-dimensional inputs such as images and massive arrays of numbers which they try to map to a single value (a class or a category) and that’s why the training process is extremely complex and long. And their performance accuracy (especially in the case of deep learning systems) is only as good as the training dataset.

Reinforcement learning is even more prone to sample inefficiency. For example, Facebook’s Go-playing model has to be trained for two weeks (by playing around 20 million games on 2000 GPU processors simultaneously) before it can perform on a close-to-human level. That’s far more than a person could play in a lifetime.

So far, RL has only shown outstanding performances in simulated environments and it’s been impractical for most real-world applications as we can’t, say, have a robot destroy millions of cars before it learns to drive adequately.

It’s becoming clear now that making algorithms smart isn’t about picking the best language for machine learning or adding layers and architectural complexities to the latest neural net models. It’s about ditching entirely the process of training with labeled data.

The post Intelligent Machine Learning Tools with Self-Organizing Maps and Other Unsupervised Methods appeared first on IT Consulting Company Perfectial.

Solving Financial Fraud Detection with Machine Learning Methods

Roksolana Kryshtanovych — Fri, 24 Apr 2020 14:30:27 +0000

For decades, financial organizations used rule-based monitoring systems for fraud detection.

These legacy solutions were deployed in SQL or C/C++. They were attempts of the engineers to transfer the knowledge of domain experts into sequel queries, which would typically end up being long, convoluted, and extremely brittle.

And whenever they tried to change parts of these fraud detection systems later, to update a threshold or something, it led to the breaking of the entire codebase.

This prevented banks from fighting fraud effectively – the criminals would just come up with new ways around alert triggers in their weak, rule-based platforms.

So now many financial firms have abandoned their legacy tools to try and solve fraud detection with new-age machine learning solutions, and more still are planning to follow suit in the future.

ML algorithms can process millions of data objects quickly and link instances from seemingly unrelated datasets to detect suspicious patterns. They’re one of the only tools left that can help banks and FinTechs keep up with new defrauding schemes, which are growing increasingly sophisticated.

But it might be unclear for someone who’s not a data scientist which algorithm to opt for to help their company identify illicit transactions. In this post, we’ll describe a few popular choices.

The post Solving Financial Fraud Detection with Machine Learning Methods appeared first on IT Consulting Company Perfectial.

How to Remotely Manage Virtual Teams in 2020

Roksolana Kryshtanovych — Tue, 14 Apr 2020 18:29:27 +0000

With new technologies coming to the workspace, remote work is becoming not only manageable but sometimes necessary – as in the recent economic situation caused by the pandemic. An economic recession or even crisis is inevitable and can significantly harm global production, supply chain and have a disruptive financial impact on firms and financial markets.

If you take a look at the economic impact of COVID-19 on China’s economy, you can see that the investment in fixed assets and retail sales are the spheres that suffered the most loss – 24.5% and 20.5%, respectively.

The economies of other countries sadly seem to be following China’s footsteps; some are now delving into an even deeper recession. So, overcoming the COVID-19 impact on the economies is a real challenge for companies now. Therefore, in today’s article, we’re going to explore how remote teams can contribute to your crisis management strategy and help keep your workflows going with minimum loss.

The post How to Remotely Manage Virtual Teams in 2020 appeared first on IT Consulting Company Perfectial.

How RPA, AI and IPA Drive Value in Banking

Roksolana Kryshtanovych — Wed, 25 Mar 2020 18:36:47 +0000

RPA may not be sophisticated or complex, but it’s growing rapidly. Software robots, basic as they might be, perform well on a certain class of tasks; they can carry out seamlessly structured, information-intensive, easily specifiable, routine work and thus free up employees for more emotive and value-adding activities.

The technology’s main drawback, however, is that it’s not generalizable or transferable. It relies on the explicit encoding of rules and configurations to generate workflows, and, therefore, SW robots can be thrown off track completely by even the slightest modifications in pre-defined process scenarios.

Intelligent process automation (IPA) is an attempt to fix this by injecting some more advanced AI capabilities into RPA solutions. It is a quickly emerging subfield of AI that concerns itself with providing support for automation of long-tail processes that span multiple systems and routines.

In this post, we’ll shed some light on RPA limitations and discuss some of the recent approaches to making the technology smarter.

RPA vs IPA

Legacy applications most huge enterprises still rely on for core operations are rigid, complex and disjointed so, typically, they have an employee responsible for transferring information from one cumbersome application to another manually (by interacting with apps’ graphical user interfaces), which is a tedious and time-consuming task.

RPA, in its conventional form, helps with this to an extent: software bots can record and repeat employees’ interactions with systems’ interfaces and thus remove some high-volume, repetitive tasks from their workloads.

What they can’t do, however, is identify on their own which actions belong to a certain routine (intra-routine learning) or determine which routines are eligible for automation in the first place (inter-routine learning).

The difference between RPA and IPA can be described by the following example. Suppose there’s an IT person who redirects each request (received as a CRM ticket) to an appropriate department. The standard RPA approach to eliminating the task would be to try and create a set of rules for a robot to follow that includes every possible scenario of the routine’s unfolding; but if the process has lots of variants (and requests usually come in various forms), this would quickly become too costly and complex an undertaking.

In IPA, however, we’d try to engineer an ML-powered bot that can identify patterns between requests and departments on its own given a sufficient number of examples (the bot would have to be trained to identify user intent beforehand.)

Teaching the robots to spot highly deterministic tasks that are ripe for automation can, therefore, help companies get rid of tedious work and reduce cycle times; it is also an essential step to making software robots more intelligent.

The post How RPA, AI and IPA Drive Value in Banking appeared first on IT Consulting Company Perfectial.

How It’s Changing the Future of Augmented Reality

Roksolana Kryshtanovych — Tue, 24 Mar 2020 15:10:02 +0000

Augmented Reality is becoming more and more popular with developers and smartphone users but it still hasn’t taken its place among omnipresent technologies. To experience AR, users have to install a dedicated application but often they delete it after a few uses or don’t even bother to download it at all. WebAR is set to become the solution for users’ reluctance to experience it via apps, and there’s some pretty good evidence that it’ll actually succeed.

What is WebAR?

WebAR is an Augmented Reality experience that is accessed via a web browser instead of an app. It provides this web-based AR experience by using technologies like WebGL, WebRTC, WebVR and APIs. In simple words, you just need a phone to access it.

WebAR enables smartphone users to discover AR technology in the easiest way— via the web— without the burden of installation. It simply breaks barriers and creates interactive 3D models one can access through QR-code or a link. For now, WebAR offers limited functionality compared to an app-based AR, but it already includes simple animations, video, and a certain degree of interactivity. Also, WebAR supports image target detection.

The post How It’s Changing the Future of Augmented Reality appeared first on IT Consulting Company Perfectial.

Product Engineering: How to Create Outstanding Software Product

Roksolana Kryshtanovych — Tue, 17 Mar 2020 19:58:29 +0000

Creating an end-to-end software product is a complicated process that involves a cycle of actions and decisions. Sometimes you have an idea for a product but no clue where to go with it. A Software Development Company, with the help of Product Engineering Services, can help you to evaluate your idea, suggest quickest implementation scenarios, and create a map for product development.

Software Product Engineering is a service that involves all stages of product creation: design, development, testing, deploying. But, the goal of Product Engineering is more challenging than simply delivering the final product – it’s to ensure that the product is functional and satisfies the needs of its end-user. Product Engineer is concerned with establishing whether the product will survive in the real world after the launch, which they determine by analyzing how it complies with the market requirements. Within the IT context, the product can be a piece of software, an app, or a business system. Product Engineering is dealing with the following specifics of the product:

Quality
Usability
Functionality
Durability
So, let’s take a closer look at who the person behind the Product Engineering Service is and what their role is in the product creation cycle.

Who is a Product Engineer?

Product Engineer is some kind of a “superhero” that has to combine the knowledge of different positions: Business Analyst and Project/Delivery Manager. This is a person that is familiar with the market and makes sure that the product complies with the consumer’s expectations. The responsibilities of the Product Development Engineer involve identifying the problems and interests of the users and creating a prototype that can offer solutions. Also, they analyze what’s trending and in-demand on the market and in the industry in general. Another concern of the Product Engineer is the budget needed for the end-to-end product creation, feasibility, and the return on investment (ROI) of the particular product idea.

Usually, the Product Engineer needs to have an excellent communication environment in order to have access to all the information they need.

The post Product Engineering: How to Create Outstanding Software Product appeared first on IT Consulting Company Perfectial.

Proximity Marketing: How it Works and Why You Need It

Roksolana Kryshtanovych — Fri, 13 Mar 2020 13:06:47 +0000

The evolving technological development often means more opportunities for businesses. This is the case with proximity marketing. This is a technology that takes the advertisement campaigns to a whole new level. Recently, the proximity marketing devices called beacons became a revelation for the retail stores and small start-ups. So, let’s take a look at what proximity marketing is and how it works.

The Concept of Proximity Marketing

Imagine you have opened a clothing store with the stock of great Scandinavian fashion. However, you do not have a base of loyal clients as your enterprise has just been born, and there are plenty of H&M’s in the district. How can you attract new clients and cope with the competition? The answer could be mobile proximity marketing.
The concept is straightforward. When the potential customer approaches your business asset, they automatically get the notification on their phone. It can be anything: new offers or an invitation to find out more about your business. What is great about this technology is that it sends the message at the right time, increasing the probability of customers actually becoming interested in the deal.
As in the modern era, people are inseparable from their mobile devices, this technology proves to be a useful addition to the firm’s advertising campaigns. According to JiWire, more than 53% of customers would not mind sharing their current location and in return receive advertising that is relevant to them at the moment.

Proximity Marketing vs Spam

One of the ways to inform potential clients about your business and great deals it can offer is spamming them via e-mail. The problem with this strategy is that people either don’t pay any attention at all to these messages or get irritated by them. The reason is that there is a significant number of spam which is often not targeted well. According to the Pew Research Center, 69% of mobile phone users get unwanted spam messages.
That is why Proximity Marketing is a great way to stand out and offer people what they actually want. Therefore, when a potential customer approaches your business asset, they get a message at the right time and right place without being annoyed.
Moreover, the proximity market technology offers a possibility to make advertising more personalized, which means that you can enhance the positive experience of your customers. This feature proves to be very effective as 88% have stated a significant increase in their business success as a result of the implementation of a personalized marketing campaign.
It is important to say that despite the fact that proximity marketing contributes considerably to a more personalized interaction with the customers, it does not breach the personal data security. The private information of the clients is not collected and only location info, history of purchases etc. are used. This makes mobile proximity marketing more trustworthy and more appealing for the customers.

The post Proximity Marketing: How it Works and Why You Need It appeared first on Software Development Company Perfectial.

Self-supervised Learning, Future of AI

Roksolana Kryshtanovych — Fri, 13 Mar 2020 13:01:00 +0000

Self-supervised learning is one of those recent ML techniques that have made waves in the data science community, but have so far been flying under the radar as far as Entrepreneurs and Fortunes of the world go; the general public is yet to learn about the concept but lots of AI folk deem it revolutionary.

The paradigm holds vast potential for enterprises too as it can help tackle deep learning’s most daunting issue: data/sample inefficiency and subsequent costly training.

We believe strongly that, as a business owner, you should acquaint yourself with this inherently complex subject and we’ll gladly help.

Hand-crafted Feature Learners and Deep Learning Systems: AI Past vs AI Future

These days, when someone mentions AI’s transformative potential, they’re likely speaking about machine learning. And when you’re reading another sensationalist post about the groundbreaking progress in ML, the authors are probably referring to deep learning (and supervised deep learning in particular).

Since human-generated features tend to be brittle, the AI community is increasingly embracing deep learning systems (for a number of tasks), which can learn data’s distinctive properties by themselves given an objective function.

These neural nets aren’t magic, though, they’re just cleverly applied statistics and algebra.

The basic build of a deep learning system includes a succession of linear and point-wise non-linear operators. Neural nets first encode every input, turning into a vector or a list of numbers, then do computations on the received values using various matrices in their neurons (which their layers consist of) and, afterward, pass the resulting vectors through a bank of nonlinear functions (such as ReLU) and a classifier at the very end of the architecture.

When the machine fails to output what is expected of it (a correct class/label in supervised learning), data scientists can tweak the network’s parameters (all the modules are trainable) until the results are satisfactory. This is done via gradient descent which, in turn, is computed by a backprop algorithm.

It turns out computers are exceptionally good at learning functions that map inputs to human-generated labels under one condition: an enormous amount of labeled data must be fed to them first. There’s also another drawback: these models, determined to classify the input into a category, don’t learn much about the inherent properties of input elements. The feedback the machine is given is scarce in supervised learning, so naturally the networks are very sample inefficient.

This creates a significant issue: high-quality data is often hard to come by for a lot of companies and obtaining an annotated dataset can prove too costly an undertaking even for large organizations.

The post Self-supervised Learning, Future of AI appeared first on Software Development Company Perfectial.