1 . JÜRGEN SCHMIDHUBER STATED:

Co-Founder & Chief Scientist of NNAISENSE, Scientific Director of Swiss AI Lab IDSIA and Professor of AI, USI & SUPSI

LinkedIn : https://www.linkedin.com/in/j%C3%BCrgen-schmidhuber-39226872/

Website : http://www.idsia.ch/~juergen/

1 . WHERE DO YOU REQUIRE THE DATA?

The data has to be collected in a source system, like SQL or Hadoop data warehouses, and 2.then fed into an analytics program like SAS, Python, R, etc. The results can then be fed back into the source data warehouses. 

2. HOW TO COLLECT DATA FOR DATA SCIENCE PROJECTS? 

I work with casinos and they collect the data when customers use player cards.. Credit card companies will collect data with every transaction a customer makes with them. Social media is a vast trove of customer data as well. 

3. WHAT ARE THE INTERNAL SOURCES, EXTERNAL SOURCES, ANY FREE DATA SOURCES?

There are plenty of free data sources, including social media sites like Facebook, Twitter, Instagram, Weibo, etc. 

4. HOW THIS DATA IS USED IN WEB SCRAPING?

Data is collected from a website and then can be used to do customer intel, i.e., what price do our products sell compared to our competitors, etc.

2 . ELIAS LANKINEN STATED:

CEO of Deepaz.ai 

Linkedin Profile : https://www.linkedin.com/in/elias-lankinen-253835182/

Website : http://deepez.ai/

1 . WHERE DO YOU REQUIRE THE DATA?

The first thing everyone start is of course problem definition. You need to have some problems to solve before you start doing anything with the data. After that you find as much data as possible that might be somehow helpful. Bigger companies have databases that contains a lot of raw data and that's the first place to go.

2. HOW TO COLLECT DATA FOR DATA SCIENCE PROJECTS?

The second place is public data sets. Then often in smaller companies it's crucial to also web scrape some data because there might not be public dataset or the company often doesn't have data on their own database. All these places are important and good sources of the data. Then you have a bunch of data from many sources so the problem is how you use it. First you start by cleaning the data. It means that you replace missing values or remove them, check that the formats match, try to find a different kind of typos, and check that there are no impossible values like price of something is negative. Cleaning often takes a lot of time and it's pretty manual. But when you have clean data it is often the best to do some feature engineering meaning that you come up with some new columns. This is again pretty manual and requires domain knowledge. In the end you have clean data and maybe some own features. This can be used to create graphs to solve the problem or train machine learning model to make predictions.

3. KRZYSZTOF SUROWIECKI STATED:

Managing Partner at Hexe Data and also an analyst that holds 16 years of experience in this field.

Linkedin : https://www.linkedin.com/in/ksurowiecki/

Company Website : https://hexedata.com

1 . WHERE DO YOU REQUIRE THE DATA?

Stages of data analysis, generally, can always be divided into three groups - data preparation - data analysis - inference and recommendations. While the first two stages are quite mechanical, the third stage is crucial, especially if we leave some inference to the algorithms.

2. HOW TO COLLECT DATA FOR DATA SCIENCE PROJECTS?

In this context, I would like to draw attention to one important stage, which is crucial from the point of view of any mechanically performed analysis. I am thinking about drawing conclusions related to "causality", i.e. the cause-effect relationship between two data sets. A common mistake in this type of automated analysis is the blind faith in the correlation coefficient (e.g. Pearson or Spearman), i.e. when we see values close to 1 or -1, it is a big temptation to blindly accept that there is a strong correlation between phenomena. It is similar to 0 and no correlation. Meanwhile, the fact that there is a correlation mathematically does not mean that it actually exists - there are known cases of correlation between the wind speed over a crop of corn in the state of X and the value of the stock index, but it really does not mean that we can build our stock strategies and this does not mean that these phenomena are related.

4. KOSTAS PARDALIS STATED:

Co-Founder and CEO at Blendo.co

Linkedin ID : https://www.linkedin.com/in/kostaspardalis/

Company Website : https://www.blendo.co/

1 . WHERE DO YOU REQUIRE THE DATA? 

Identifying the needs of the market, Blendo is built supporting any kind of data source. Today a company has data in file systems, on cloud applications, on cloud-based documents and internal systems like databases. Blendo supports all of them.

2. HOW TO COLLECT DATA FOR DATA SCIENCE PROJECTS? 

Blendo allows you to connect on all the different sources, shape them in a way that is consistent for a data lake or data warehouse solution and makes sure that quality and privacy requirements are met. When you have all the data available in a repository where a data scientist can easily access them, then it’s easy for them to iterate faster and safer on any kind of data science project.

3. WHAT ARE THE INTERNAL SOURCES, EXTERNAL SOURCES, ANY FREE DATA SOURCES. 

Internal sources are usually either cloud-based file systems like AWS S3, database systems like an Azure MS SQL database and streaming data coming from a Kafka cluster or Amazon Kinesis.

External sources are usually cloud applications that a company uses, like a SalesForce CRM and Zendesk ticketing system or even a Google sheet. Free data sources are usually publicly available datasets that are preprocessed and can be used by the data scientist to enrich the data they are using. For example, any data set that is coming from a government agency.

4. HOW THIS DATA IS USED IN WEB SCRAPING

Web Scraping is just another way for generating data sets, some data might leave only on web pages and in this case we need to scrape it and restructure the data into a structured format that can be queried. The most typical case is of course Google. Google is scraping pages and makes them searchable based on keywords.

5. ALEX BEKKER STATED:

Head of Data Analytics Department at ScienceSoft, an IT consulting company

Twitter ID : https://twitter.com/alexlbekker?lang=en

Company Website :

https://www.scnsoft.com/experts/alex-bekker

1 . HOW DO YOU EXPLAIN THE DIFFERENT STAGES IN DATA ANALYTICS, AND WHERE DO YOU FIT IT IN AI AND ML ACTIVITIES?

“To describe the stages of data analytics maturity, we, at ScienceSoft, rather use the term ‘data analytics types’, and distinguish 4 of them: descriptive, diagnostic, predictive, and prescriptive. AI and ML activities (thanks to their power in identifying complex dependencies in data) belong to the advanced types: predictive and prescriptive.”

If you’d like to get more info about each type of analytics or read real-life examples of predictive and prescriptive analytics in action, you’re welcome to consult my article ‘4 Types of Data Analytics to Improve Decision-Making’.

2 . HOW TO COLLECT DATA FOR DATA SCIENCE PROJECTS ? WHAT ARE THE INTERNAL SOURCES, EXTERNAL SOURCES, ANY FREE DATA SOURCES.

“For data science projects, companies can use both internal and external data sources. For example, a deep neural networks (DNN) that is to predict an optimal inventory level can perfectly do with historical sales data, while a DNN that is to recognize medical images can be trained based on publicly available image data sets.”

After our experts preached their valuable information, it’s time to get deeper insights on the main topics they covered surrounding data.

Data analytics is the process of identifying solutions form the given data for the growth of businesses in a healthier manner. The information about a topic is right in front of you, a thorough examination is conducted out of the information, solutions, predictions, and conclusions are formed so that the purpose of the process is conducted.

There are 5 stages in a data analytics process:

1. WHY YOU NEED DATA ANALYSIS?

The first stage in data analysis is to identify why do you even need to use this solution? What is the main agenda you want to achieve with this process? As you are aware that data analytics helps to find outcomes that can help brands to grow better, you need to identify first which problems you want a solution? Id it your low revenue count which you want a solution or how can your lead conversion increase better? When you have the problems in front of you, it becomes easier to understand how conducting a data analysis activity can help.

2. COMMENCE COLLECTION OF DATA FROM VARIOUS SOURCES

The next stage is to take the purpose of the first step and start finding relevant data that can be useful in conducting the data analytic process. Start finding solutions from sources such as social platforms, CRM or other places where the data is more relevant and can help you. For instance, you may want to know how you can increase lead conversion rates, so you would search through the market conditions and see what the market is expecting, you would also get information from your leads itself via reviews, feedback and many more. Get as much information you can from various sources so that it becomes easier to extract a suitable outcome.

3. FIND YOUR WAY THROUGH UNNECESSARY DATA

Since your now extracting data from various sources, the third step is to scrape through all the bad data you find. Not all the data you collect are going to be useful to serve your purpose. You must have only clean data that is relevant and useful to you. Hence this is one of the most crucial parts in the data analytic process.

4. START ANALYZING THE DATA IN HAND

The next stage is to begin analyzing the clean you have in hand. Using measures such as predictive analysis, data mining and several other factors it becomes easier to study the information you have in hand. With such measures, charts, graphs, and conclusions can be easier to identify and help make the right decision of certain scenarios.

5. INTERPRET THE OUTCOMES RECEIVED AND BEGIN APPLYING THEM TO REAL-TIME SCENARIOS

In the last stage you then start applying the outcomes you retrieve from the data collected. The results might not be accurate but you can conduct experiments to understand what works best. For instance, to solve the lead conversion rates problem, solution A states to grow your social presence more and solution B states that you need to improvise on your branding strategy. The experiment will continue and which one will work the most will continue to apply.

Data analytics is beneficial for those businesses that practice it. The reason being:

• It preps brands to improvise their solutions accordingly
• It helps brands to capture leads at the right time
• It offers a great advantage for brands to increase their revenue via conducting the methods which will yield them success numbers
• It carves an image of helping brands to conduct their lead activities more efficiently
• Saves time and effort to help brands decide what measures will work during lead driven activities
• It helps brands to always stay on top when it comes to creating a lead driven strategy

Data analytics requires valuable data that can help conduct this process. But with multiple strict restrictions to access data online, how can you conduct data analytics that can grow your business better?

Web scraping can help you get the data your brand needs to successfully comprehend its future growth. Web scraping is the process of extracting data from a website or other source of information and saving that data in your system in the format you wish to see it, for instance, CSV file.

Web scraping is becoming popular because all you need to do is:

Click on the website you wish to scrape

Once the website is chosen web scraping commences its job of extracting that valuable data

The information will then be saved directly in your system

Web scraping scrapes through any information within minutes and allows you to easily monitor and take charge thereafter. When the data you need is in your hand, conducting data analytics becomes much easier and more enhanced. This allows you to find solutions that will work since the data retrieved is a valuable piece of information.

Since web scraping conducts multiple or even regular data scraping activities, chances are at times you can lie at risk of being caught or blocked. Since the action is repetitive you do stand under the risk of getting caught. Hence, to avoid this you must use a secure platform while this process is being conducted.

This is when proxy servers can help you. 

Proxy servers are middlemen who protect and ensure that any information being received is free from risks or any other harm. For instance, say you want to view a website, you send a request to access it. Instead of going to the website the request will be sent to the proxy server first who changes your IP address. 

When that happens, the proxy server will send the request. Once the website approves the request you’ve sent, the access will be shared. The access will be sent to the proxy server first to ensure that there is no harm and then it is sent to you.

The main reason why you get blocked when viewing a website is that your IP address gives way. Proxy servers ensure that all your IP address gets blocked so that you can conveniently scrape through any information you require.

Since you now have all the solutions you need to practice data analytics did you know that the influence of artificial intelligence and machine learning incorporate it as well?

AI’s impact on marketing is growing, predicted to reach nearly $40 billion by 2025. 

AI works better with data analytics. As data analytics finds outcomes to solutions that can help brands achieve better brand growth AI ensures that this process conducted with sheer accuracy.

Interesting Read : Difference Between Big Data, Data Science and Data Analytics

AI conducts data analytics without the need of any human assistance. With the help of algorithms, AI creates assumptions, evaluates data much quicker with the information it has in hand.

Evaluating huge amounts of data is a tiresome task and with the market conditions changing all the time, this process must be conducted at a much quicker pace which is exactly why AI is helpful here. 

Forbes has explained a brilliant example of how useful AI can be:

‘’AnyBank's credit card loyalty program could utilize machine learning to determine that 1,000 of its male members live near a golf course, have not golfed before but enjoy sports. It also determines that many women members in the loyalty program are equally likely to be interested. It also sets parameters for the golf season in certain climate zones, such as the Southern U.S. It further determines a microsegment to offer Saturday afternoon to men and women without young children, who can more likely take the time on a Saturday. The system also acknowledges other sports-loving middle-aged men and women who live near a golf course and have young children, which would prompt other, more family-oriented offers.’’

Machine learning is also another aspect that fits in data analytics. The difference between machine learning and AI is while AI can create assumptions and continue to test and grow in terms of learning, machine learning, on the other hand, indulges in identifying patterns that can be much easier to create an outcome without the need for any human supervision.

Machine learning is a branch of AI and in data analytics where a human can take time to identify data and then come up with an expected outcome, machine learning uses systems which help to draw out a pattern which can not only enhance inaccurate decisions but also speed things up a notch for better solution planning.