😎💥Cool AI services for working with Big Data
AskEdith - works in a “Data Chat” mode where you can connect to your data sources and ask questions in natural language. AskEdith provides answers in a variety of formats, including visualizations. Capable of analyzing large volumes of data and connecting to various databases and CRM systems such as Google Sheets, Airtable, PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery, Redshift and others.
Tomat.AI - allows you to analyze large CSV files without the need for programming or creating a special formula. You can easily filter, sort, merge multiple files and create visualizations in just a few clicks.
Coginiti - Creates SQL queries, explains their meaning and, if necessary, optimizes performance. Supports access to various data warehouses, Redshift, Microsoft, Snowflake, IBM, BigQuery, Yellowbrick, Databricks and others. In addition, there is a repository for storing SQL queries.
Formula God - Artificial intelligence that integrates into Google Sheets, including working with data without the need to use formulas. You can create queries using normal spoken language.
Simple ML for Sheets – With these extensions, anyone can use machine learning in Google Sheets rather than define programming and machine learning. But it is also suitable for experts if they want to quickly select a task for small amounts of data. Developed by the TensorFlow Solution Forest team.

📝🤔📊 One Hot Encoding: advantages and disadvantages
One Hot Encoding (OHE) is a method for representing categorical data as binary vectors. This method is widely used in machine learning to work with data that contains categorical features, that is, features that are not numeric. With One Hot Encoding, each category is converted into a binary vector where all values are zero except one, which corresponds to the category of a given feature.
Advantages of One Hot Encoding:
1. Suitable for Machine Learning Algorithms: Many machine learning algorithms such as linear regression, decision trees and neural networks work with numerical data. One Hot Encoding allows you to convert categorical features into numbers, making them suitable for analysis by algorithms.
2. Useful for categorical features without ordered values: If categories do not have a natural order or are unevenly distributed, One Hot Encoding may be a preferable representation method over Label Encoding.

Disadvantages of One Hot Encoding:
1. Data dimensionality: Transforming categorical features with a large number of unique categories can result in a significant increase in data dimensionality, which can degrade the performance of machine learning algorithms and require more memory.
2. Multicollinearity: When you have multiple categorical features with a large number of unique categories, multicollinearity problems can arise, where one feature is linearly dependent on the others. This can make the models difficult to interpret.
3. Increasing computational complexity: Increasing the data dimensionality can also lead to an increase in model training time and a more complex feature selection task.

Thus, the choice between One Hot Encoding and other categorical feature encoding methods depends on the specific task and the machine learning algorithm you plan to use.

🌎TOP DS-events all over the world in October
Oct 4-5 - Chief Data & Analytics Officers - Boston, USA - https://cdao-fall.coriniumintelligence.com/
Oct 10-11 - CDAO Europe- Amsterdam, Netherlands - https://cdao-eu.coriniumintelligence.com/
Oct 14-16 - International Conference on Big Data Modeling and Optimization - Rome, Italy - http://www.bdmo.org/
Oct 16-20 - AI Everything 2023 - ОАЭ, Дубай - https://ai-everything.com/home
Oct 16-19 - The Analytics Engineering Conference - San Diego, CA, US - https://coalesce.getdbt.com/
Oct 18-19 - Big Data & AI Toronto - Toronto, Canada - https://www.bigdata-toronto.com/
Oct 23-26 - International Data Week 2023 - Salzburg, Austria - https://internationaldataweek.org/idw2023/
Oct 24-25 - Data2030 Summit 2023 - Stockholm, Sweden - https://data2030summit.com/
Oct 25-26 - MLOps World - Austin TX - https://mlopsworld.com/

⚔️🤔Greenplum vs Hive: advantages and disadvantages
Greenplum and Hive are two different data science technologies used in the field of big data and analytics.
Greenplum benefits:
1. High performance: Greenplum provides a multi-user analytics engine with a distributed architecture. This enables fast query processing and aggregation performance, making it an excellent choice for real-time analytics.
2. Scalability: Greenplum is designed to scale horizontally. You can easily add new nodes to increase performance and storage as needed.
3. Data Management: Greenplum provides tools for data management, including replication, backup and monitoring, making it more suitable for business needs that require data reliability and availability.
Disadvantages of Greenplum:
1. Challenging Setup: Installing and configuring Greenplum can be a challenging task. Requires experience and knowledge of system architecture for optimal performance.
2. Not suitable for all use cases: Greenplum is best suited for analytical tasks and storing structured data, but is not the optimal choice for processing semi-structured and unstructured data.

Benefits of Hive:
1. Easy to use and configure: Hive is built on top of Hadoop and provides an SQL-like interface for querying data. This makes it more accessible to analysts and developers without big data experience.
2. Compatible with Hadoop: Hive is integrated with Hadoop and can use it for data storage and processing. This makes it a good choice for projects using Hadoop.
3. Support for a variety of data formats: Hive supports various data formats including JSON, Parquet, Avro and others, making it convenient for analyzing a variety of data.
Disadvantages of Hive:
1. Poor performance: Hive is slower than Greenplum due to the fact that queries are translated into MapReduce tasks, which can lead to significant delays.
2. Limited support for complex analytic queries: Hive is not as well suited for running complex analytic queries as Greenplum due to its limited query optimization capabilities.
3. Not suitable for real-time: Hive is best suited for batch data processing and is not a suitable choice for real-time analytics.

📖📚Selection of books for data analysts
Анализ данных с помощью Python - a complete guide to the science of data, analytics and metrics with Python.
Математика для машинного обучения - the book discusses the basics of mathematics (linear algebra, geometry, vectors, etc.), as well as the main problems of machine learning.
Интерпретируемое машинное обучение - a guide to creating explainable black box models
Понимание статистики и экспериментального дизайна -this textbook provides the basics necessary for the correct understanding and interpretation of statistics.
Этика и наука о данных - in this book the author introduces us to the principles of working with data and what to do to implement them already Today
Использование науки о данных в здравоохранении - the book discusses the use of information technology and machine learning to combat diseases and health promotion

📊📉📈User analysis using a dataset from Yandex
Yandex makes the largest Russian-language dataset publicly available reviews of organizations published on Yandex Maps. It contains about half a million user reviews of various organizations, collected in January-June 2023.
Dataset features:
500,000 unique reviews
Texts are cleared of personal data (phone numbers, email addresses)
The dataset does not contain short monosyllabic reviews
Quite recent reviews: from January to July 2023

📋💡🔎Selection of datasets for NLP
КартаСловСент — words and expressions equipped with a tonal label (“positive”, “negative”, “neutral”) and a scalar value of the strength of the emotional-evaluative charge from the continuous range [-1, 1].
WikiQA is a set of pairs of questions and proposals. They were collected and annotated to explore answers to questions in open domains
Amazon Reviews dataset - this dataset consists of several million Amazon customer reviews and their ratings. The dataset is used to enable fastText to learn by analyzing customer sentiment. The idea is that despite the huge volume of data, this is a real business problem. The model is trained in minutes. This is what sets Amazon Reviews apart from its peers.
Yelp dataset is a set of businesses, reviews and user data that can be used in a Pet project and scientific work. Yelp can also be used to train students while working with databases, learning NLP, and as a sample of manufacturing data. The dataset is available as JSON files and is a “classic” in natural language processing.

⚔️⚡️Altair vs. Matplotlib: advantages and disadvantages of Big Data visualizations
Matplotlib is an old-timer in the world of data visualization and is widely used in the Python community.
Advantages of Matplotlib:
1. Maximum flexibility: Matplotlib allows you to create almost any kind of plot and customize every detail. You can create static and animated graphics suitable for various purposes.
2. Large Community and Documentation: Due to its popularity, Matplotlib has a huge user community and extensive documentation. This makes it a great choice for beginners and experienced users.
3. Wide variety of graphical elements: Matplotlib provides a rich selection of graphical elements such as lines, points, columns, and more, allowing you to create a variety of plots.
Disadvantages of Matplotlib:
1. Complexity: Creating complex plots with Matplotlib can be non-trivial and require significant effort and code.
2. Default Appearance: Plots created with Matplotlib may not look very attractive by default, and often require additional and quite time-consuming work to improve them.

Altair is a newer library that aims to make it easier to create declarative graphs.
Altair advantages:
1. Declarative approach: Altair offers a declarative approach to creating graphs, which means you describe what data you want to visualize and how, and the library takes care of the details.
2. Ease of Use: Altair allows you to create beautiful graphics with minimal code. This makes it a great choice for rapid prototyping and beginners.
3. Pandas Integration: Altair integrates well with the Pandas library, making it easy to work with data.
Disadvantages of Altair:
1. Limited customization options: Compared to Matplotlib, Altair provides fewer options for customizing plots. If you need complex and non-standard graphics, this may be a limiting factor.
2. Smaller community and documentation: Altair, being a new project, has a smaller user community and less extensive documentation.
The choice between Altair and Matplotlib depends on your specific needs and experience level. Matplotlib is suitable for those who need complete flexibility and control over their plots, while Altair provides a simple and declarative way to create beautiful plots with minimal effort.

😎⚡️Visualizing astronomy is now even easier in Python
APLpy - the Astronomical Plotting Library in Python) is a Python module designed to create astronomical image publishing plots in FITS format.
Have you ever wanted to try astronomical data visualization? Now this can be easily done in Python using this library. To install this library you just need to run the following command:
pip install aplpy

📊📝Agricultural data of the European Union is publicly available

EuroCrops is a comprehensive collection of datasets that brings together all publicly available agricultural data in the European Union.
This project is funded by the German Space Agency DLR on behalf of the Federal Ministry of Economic Affairs and Climate Change.

⚔️📊LDA vs t-SNE: advantages and disadvantages
Two popular methods for data analysis, LDA (Linear Discriminant Analysis) and t-SNE (t-Distributed Stochastic Neighbor Embedding), are used to solve various problems. They both have their own unique advantages and disadvantages. Let's take a closer look at them.

Advantages of LDA:
1. Classification: LDA is designed for classification and data partitioning tasks. It aims to maximize the distance between classes, making it an excellent choice for classification and pattern recognition problems.
2. Interpretability: LDA creates new features (linear combinations of the original ones) that can be interpreted as “discriminant axes”. This makes it easier to explain how and why data is shared.
3. Efficiency on large data: LDA is generally more efficient when dealing with large amounts of data than t-SNE. It may be faster and require less memory.
Disadvantages of LDA:
1. Linear nature: LDA assumes that data is linearly separable, which may limit its applicability in problems where classes cannot be linearly separable.
2. Lack of visual information: LDA creates a new feature space, but does not necessarily preserve the similarity between given points. This makes it less suitable for data visualization.

Advantages of t-SNE:
1. Robust to non-linear relationships: t-SNE can detect non-linear relationships in data, making it a good choice for data visualization in cases where linear separation is not sufficient.
2. Displaying high-dimensional data: t-SNE can deal with high-dimensional data while preserving its structure while reducing dimensionality.
3. Better Visualization: t-SNE provides more visualization of data by grouping similar points into dense clusters.
Disadvantages of t-SNE:
1. Sensitivity to parameters: The choice of parameters such as perplexity can greatly affect the results of t-SNE. A thorough analysis of the parameters is necessary.
2. Computational complexity: t-SNE can be computationally expensive and slow when dealing with large data sets.
3. Lack of interpretability: Since t-SNE strives for visual grouping of points, it does not create interpretable new features.

Thus, the choice between LDA and t-SNE depends on the specific goals of the analysis. LDA is better suited for classification and interpretability tasks, while t-SNE is generally preferred for visualization and detection of nonlinear relationships.

🤔Data tagging: advantages and downsides
Data tagging is the process of assigning labels or annotations to specific elements in a data set to train machines to understand and extract information from that data. Data labeling plays an important role in machine learning, deep learning, and data mining because it allows algorithms to understand which objects or factors in the data are important and which are unimportant.
Benefits of data tagging:
1. Improve model accuracy: Data labeling helps create more accurate and reliable models because algorithms can learn from the correct labels and avoid errors.
2. Training algorithms: Labeled data allows machine learning algorithms to be trained more efficiently, making them capable of solving complex problems such as pattern recognition, text classification, forecasting and others.
3. Expanding the functionality of applications: Labeled data allows you to develop more intelligent applications and services, such as virtual assistants, automated systems and much more.
Disadvantages of data tagging:
1. Resource-intensive: Data tagging requires significant effort and resources, especially when it comes to large data sets or complex tasks.
2. Subjectivity: Data labeling may depend on the subjective judgments of the labelers, which can lead to errors and inaccuracies.
3. Task limitation: Data labeling is limited to a specific training task, and changing this task may require re-labeling the data.
4. Updating Data: Labeled data can become outdated over time, and the labeling needs to be updated periodically to keep models up to date.

Overall, data labeling is an integral part of many machine learning projects, and its benefits often outweigh its disadvantages, especially when the labeling process is properly organized and managed.

🌎TOP DS-events all over the world in November
Nov 9 - Big Data Analytics & AI - London, UK - https://whitehallmedia.co.uk/bdanov2023/
Nov 14-15 - SLAS-FHNW 2023 Data Sciences and AI Symposium - Basel, Switzerland - https://www.slas.org/events-calendar/slas-2023-data-sciences-and-ai-symposium/
Nov 20-21 - Gartner IT Infrastructure, Operations & Cloud Strategies Conference - London, UK - https://www.gartner.com/en/conferences/emea/infrastructure-operations-cloud-uk
Nov 20-24 - THE BIGGEST AI EVENT WORLDWIDE - Belgrade, Serbia - https://datasciconference.com/
Nov 21-24 - BIG DATA CONFERENCE EUROPE - Vilnius, Lithuania - https://bigdataconference.eu/
Nov 23-24 - Data Science Summit - Warsaw, Poland - https://dssconf.pl/en/
Nov 25-26 - 4th International Conference on Data Science and Applications - London, UK - https://www.cndc2023.org/dsa/index
Nov 27-29 - THE GLOBAL BIG DATA ANALYTICS IN POWER & UTILITIES INDUSTRY FORUM - Berlin, Germany - https://berlin-energy-summit.com/etn/the-global-big-data-analytics-in-power-utilities-industry-forum-27-28-29-november-2023/
Nov 30 - Dec 1 - AI & Big Data Expo Global - London, UK - https://www.ai-expo.net/global/speakers/

📝🔎💥Data temperature management is now even easier
Great Expectations (GX) is an open source Python-based tool for data quality control. It provides a team of data scientists with the ability to analyze and validate data, as well as create reports using it. This tool has a user-friendly command line interface (CLI) that allows you to create new tests and edit existing sources. It's important to note that Great Expectations can be integrated with a variety of data extraction, transformation, and loading (ETL) tools such as Airflow and various database management systems. A complete list of related integrations and official documentation can be found on the Great Expectations website

📝📚Selection of books on Data Mining
Data Mining: Practical Machine Learning Tools and Techniques - The book provides an introduction to the fundamentals of Data Mining and uses the popular Weka tool to train machine learning algorithms
Introduction to Data Mining - a classic book that covers the basic concepts and techniques of Data Mining
Principles of Data Mining - this book provides an extensive discussion of the principles and methods of Data Mining
Data Mining Techniques: For Marketing, Sales, and Customer Relationship Management - the book is focused on the use of Data Mining in marketing and customer relationship management
Data Science for Dummies - a good option for beginners, the book covers many topics including Data Mining, machine learning and data analysis

🤖⚡️🔎Selection of AI-based services for Big Data analysis
AskEdith - Simplifies data analysis by allowing users to ask questions and get instant information. Expands the capabilities of “self-service analytics” by providing secure and reliable access to data. Compatible with all databases and CRMs (Google Sheets, Airtable, PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery and Redshift, etc.)
Tomat.AI - An artificial intelligence-powered tool that allows data scientists to easily explore and analyze large CSV files without the need for coding or writing formulas. You can open and view huge CSV files with just a few clicks
Coginiti - Allows users to generate SQL queries using natural language hints, optimize existing SQL queries, explain common SQL in an integrated catalog, provide detailed explanations and solutions to errors, and explain plans query execution for better optimization. The AI assistant continually evolves based on every interaction, tailoring recommendations and suggestions to suit individual needs
Speak Ai - A language data analysis and research platform that offers trannoscription, data mining, and sentiment analysis capabilities for various media types. It allows automatic trannoscription, bulk analysis, visualization and data collection for use in research, market analysis and competitive analysis. The tool also offers a shared media repository, an AI-powered text hint system, and a SWOT analysis solution, among other features
Formula God - An artificial intelligence tool built into Google Sheets. It uses artificial intelligence to help users manipulate and calculate data across a full range of cells
Simple ML for Sheets - useful for machine learning experts who want to quickly iterate or prototype on small (e.g. <1 million examples) tabular data sets. Simple ML for Sheets is a Google Sheets add-on from the TensorFlow Decision Forests team.

📝🔎Apache Flink: advantages and disadvantages
Apache Flink is a distributed real-time data processing system that provides capabilities for streaming data processing and real-time analysis.
Benefits of Apache Flink:
1. Stream Data Processing: Flink is designed to efficiently process data in real time, allowing you to quickly respond to changes and events.
2. High Performance: Flink provides high performance through optimized query execution and efficient task distribution across the cluster.
3. Flexibility and Scalability: Flink provides flexibility in defining and modifying stream computing. Also worth noting is the increase in performance as the volume of processed data increases.
Disadvantages of Apache Flink:
1. Complexity of Setup: Setting up and managing an Apache Flink cluster can require significant effort and experience.
2. Lack of widespread popularity: Compared to some other real-time data processing systems, Apache Flink is not as widely used, which may affect the availability of resources and the support community.
3. Integration Challenges: Integrating Apache Flink with existing systems and tools can be challenging, requiring data reworking to be compatible with other systems' formats and structures.
Overall, Apache Flink provides powerful real-time data processing capabilities, but requires careful implementation and management to achieve maximum performance and reliability.

💥😎Selection of open datasets for various areas
This collection is a list of high-quality open datasets for machine learning, time series, NLP, image processing, etc., focused on specific topics.
Datasets are available at this link

📝A little about ClickHouse: advantages and disadvantages
ClickHouse is an open source columnar database designed for processing analytical queries with large volumes of data.
Advantages of ClickHouse:
1. High performance: ClickHouse is optimized for running analytical queries on large volumes of data. It provides high query speed due to its columnar data structure and other optimizations.
2. Scalability: ClickHouse easily scales horizontally, allowing you to add new cluster nodes to process a growing volume of data.
3. Efficient use of resources: Thanks to columnar layout and data compression, ClickHouse can efficiently use storage resources, which reduces disk space consumption.
4. Low read overhead: Thanks to its data structure and optimizations, ClickHouse provides high read performance.
Disadvantages of ClickHouse:
1. Limited transaction support: ClickHouse is focused on analytical queries and does not have full transaction support, which can be a disadvantage for applications that require strong data consistency.
2. Limited write support: ClickHouse is designed primarily for reading data, and write operations may be less efficient than other database management systems for large change volumes.
3. Insufficient indexing support: ClickHouse has limited indexing support compared to some other DBMSs, which can affect the performance of search operations.
4. Difficult to maintain and set up: Setting up ClickHouse may require some skill and understanding of its architecture, which may make it less attractive to less experienced administrators.
Overall, the choice of ClickHouse depends on the specific needs of the project. If your tasks involve analytics and processing large volumes of data, ClickHouse may be an excellent option. However, if highly consistent transactions and writes are required, other solutions may be worth considering.

😎🔎Selection of useful OLAP services for processing Big Data
Apache Druid is a real-time OLAP engine. It is focused on time series data, but can be used for any data. It uses its own columnar format that can highly compress data, and it has many built-in optimizations such as inverted indexes, text encoding, automatic data folding, and more.
Apache Pinot - Offers lower latency thanks to the Startree index, which does partial precomputation, so it can be used for user-facing applications (it was used to fetch LinkedIn feeds). This uses a sorted index instead of an inverted one, which is faster.
Apache Tajo - Designed to perform ad hoc queries with low latency and scalability, online aggregation and ETL for large data sets stored in HDFS and other data sources. It supports integration with Hive Metastore to access shared schemas.
Solr is a very fast open source enterprise search platform built on Apache Lucene. Solr is robust, scalable, and fault-tolerant, providing distributed indexing, replication and load-balanced queries, automatic failover and recovery, centralized configuration, and more.
Presto is an open source platform from Facebook. It is a distributed SQL query engine for running interactive analytical queries against data sources of any size. Presto lets you query data where it lives, including Hive, Cassandra, relational databases, and file systems. It can query large data sets in seconds. Presto is independent of Hadoop, but integrates with most of its tools, especially Hive, to run SQL queries.

🌎TOP DS-events all over the world in December
Dec 4-5 - ICDSTA 2023: 17 - Tokyo, Japan - https://waset.org/data-science-technologies-and-applications-conference-in-december-2023-in-tokyo
Dec 6-7 - The AI Summit New York - New York, USA - https://newyork.theaisummit.com/
Dec 6 - DSS NYC: Applying AI & ML to Finance & Technology - New York, USA - https://www.datascience.salon/newyork/
Dec 7-8 - ADSN 2023 Conference - University of Adelaide, Australia - https://www.australiandatascience.net/event/2023-adsn-conference/
Dec 8-10 - CDICS 2023 - Online - https://www.cdics.org/
Dec 11-15 - DSWS-2023 - Tokyo, Japan - https://ds.rois.ac.jp/article/dsws_2023
Dec 25-26 - ICVDA 2023: 17. International Conference on Vehicle Data Analytics - France, Paris - https://waset.org/vehicle-data-analytics-conference-in-december-2023-in-paris