You are on page 1of 16

[Document title]

[Document subtitle]
Table of Contents
1. Introduction:............................................................................................................................................2
Task 1:..........................................................................................................................................................2
Example SQL Code:..................................................................................................................................4
Business Implications:..............................................................................................................................6
Value Proposition:...................................................................................................................................7
Task:02.........................................................................................................................................................8
Part 1...........................................................................................................................................................8
1. Accessing external sources of data including data lakes..................................................................8
2. Current Features and Capabilities:.......................................................................................................9
3. Challenges and Considerations:...........................................................................................................9
Part 2.........................................................................................................................................................10
1. Model Explainability Using SQL in BigQuery ML: A Comprehensive Appraisal...............................10
2. Utilizing the ML.EXPLAIN Function:...................................................................................................10
3. Feature Importance Analysis:........................................................................................................11
4. Querying Explanatory Information....................................................................................................11
6. Visualizing Explanatory Data:.............................................................................................................11
8. Industry Adoption and Best Practices:...............................................................................................12
9. Continuous Improvement and Updates:............................................................................................12
10. Conclusion:..........................................................................................................................................13
10. References:..........................................................................................................................................13

1
1. Introduction:
This report looks in-depth at the process of making a machine learning model using Google's
BigQuery ML. Its main goal is to find out how well it works when you make training and checkup
tables, develop ones that follow straight lines called linear regression models, guess answers
and learn more about what BigQuery ML can do with SQL support for things we went into
depth on before talking about The aim is to check if the platform works well with data from
outside, what it can do right now and how good it is at giving clear views into guessing done by
models. They started by making two tables, training_data and evaluation_data, from the
original set of information. The tables had their data divided, with 80% used for learning and
the other 20% kept for checking. This step was made to make sure the machine learning
model's training and testing were done on different sets of data. This would help get true
results when checking how well it works. The trained model was then used to guess on the
results data table. This step was meant to test how well the model worked on fresh, never-
before-seen data. We looked at how BigQuery ML can get data from outside sources, like lakes
of info. The review of the literature showed how BigQuery changed to perfectly add machine
learning into its environment that uses SQL. It talked about the problems with putting in data
from outside sources. It focused on how it affects how well models work and their ability to
apply across different situations. Real-life examples and stories showed how companies use
BigQuery ML to get information from outside data and make smart choices. Looking at how
clear a model in BigQuery ML is, we checked tools like the EXPLAIN function for machine
learning (ML), looked into what parts of data are important to explain things well with queries
and pictures. The report talked about the good parts of using SQL for understanding what a
model is doing. It also covered its job in finding which features are most important, and being
able to show data that helps explain things better.
Task 1:
Step 1: Creating Training and Evaluation Tables:
Make two tables, training_data and testing_data, from the first set of data. The training data
table helps a computer style model learn, and the assessing results table is used to measure

2
how well that model works. Rows are chosen using age in pregnancy. Then, 80% of them is
used to learn and the rest or 20%, for checking how well it works.

Step1; Creating tables


CREATE TABLE `complete-sector-409718.countrynationality.training_data` AS
SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
Ave_OE_Gestational_Age_Wks AS label
FROM
`complete-sector-409718.countrynationality.country-nationality`
WHERE?
Ave_OE_Gestational_Age_Wks IS NOT NULL
AND RAND() < 0.8; -- 80% for training
CREATE TABLE `complete-sector-409718.countrynationality.evaluation_data` AS
SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
Ave_OE_Gestational_Age_Wks AS label
FROM
`complete-sector-409718.countrynationality.country-nationality`
WHERE
Ave_OE_Gestational_Age_Wks IS NOT NULL
AND RAND() >= 0.8; -- 20% for evaluation

Step:02
Make a gestational age prediction model called `gestational_age_prediction_model` using the training
data table. Choose the features that need to be put in (like where you live) and what we want from
them - it's called Ave_OE_Gestational_Age_Wks. Train the model to understand how features link
with target variable.
Step 2; -- Step 2: Create and Train the Model
CREATE OR REPLACE MODEL `complete-sector-409718.countrynationality.gestational_age_prediction_model`

3
OPTIONS(model_type='linear_reg') AS
SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
label
FROM
`complete-sector-409718.countrynationality.training_data`;
Step 03:
Used the trained model to guess about the evaluation_data table. Get both the real and guess age
values, along with other important traits to check them out. The guesses help measure how correct
and good the model works on new, not-seen data.
SELECT * FROM ML.PREDICT(MODEL `complete-sector-409718.countrynationality.gestational_age_prediction_model`,

( SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
label
FROM
`complete-sector-409718.countrynationality.evaluation_data`
))
Example SQL Code:

-- Step 1: Creating Training and Evaluation Tables

-- Creating training_data table


CREATE TABLE `project.dataset.training_data` AS
SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
Ave_OE_Gestational_Age_Wks AS label
FROM

4
`project.dataset.original_dataset`
WHERE
Ave_OE_Gestational_Age_Wks IS NOT NULL
AND RAND() < 0.8; -- 80% for training

-- Creating evaluation_data table


CREATE TABLE `project.dataset.evaluation_data` AS
SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
Ave_OE_Gestational_Age_Wks AS label
FROM
`project._dataset.original_dataset`
WHERE
Ave_OE_Gestational_Age_Wks IS NOT NULL
AND RAND() >= 0.8; -- 20% for evaluation

-- Step 2: Create and Train the Model

-- Creating and training the linear regression model


CREATE OR REPLACE MODEL `project.dataset.gestational_age_prediction_model`
OPTIONS(model_type='linear_reg') AS
SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
label
FROM
`project.dataset.training_data`;

-- Step 3: Making Predictions

-- Making predictions on the evaluation_data table


SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,

5
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
label,
predicted_label AS predicted_gestational_age
FROM
ML.PREDICT(MODEL `project.dataset.gestational_age_prediction_model`,
(SELECT
County_of_Residence,
County_of_Residence_FIPS,
Ave_Age_of_Mother,
Ave_Birth_Weight_gms,
Ave_Pre_pregnancy_BMI,
label
FROM
`project.dataset.evaluation_data`
)
);

Business Implications:

In this case, the ML model is trying to guess how long a pregnancy will last. It does this by
looking at things like mom's age and where she lives, along with baby weight before birth and
BMI (Body Mass Index) before becoming pregnant. This model's impact on business is wide-
reaching, especially in the healthcare industry. By using the model's guesses, doctors can make
prenatal care better with personal plans. This is done by making sure they know how long a
baby has been growing inside its mother clearly. This makes sure that pregnancy help and care
are given when needed, improving overall the health of moms and babies. The model's
predictions can be very important in deciding how to use resources at hospitals. Hospitals can
guess what moms-to-be need in different places, helping them use resources well. This ability is
very important for making sure health care places are ready if they need more help with
prenatal services. This makes the care better overall. On a larger level, the model's forecasts
can help with public health planning for all people. Health departments can use this
information to make plans and carry out specific actions, like teaching people about it through
campaigns or programs. The model's predictions take into account things such as where a
woman lives, her age when she has kids, birth weight and body size before getting pregnant.
These suggestions are very helpful for making plans to help mothers health by considering

6
regions specifically. Insurance companies can include the predictions of age during pregnancy in
their risk assessment plans. This brings a fresh aspect to making decisions about insurance. This
lets them give better guesses on how much healthcare for pregnancies costs. This helps
insurance companies make smart choices about what they cover and the cost of it. The model's
forecast helps to better understand the risks linked with mother health. This is good for both
insurers and those covered by policies. Beyond just practical use, the model's understanding of
when babies are born based on things like where a mom lives, her age at birth and other details
helps to provide important information for studies and making laws. Researchers and policy
makers can use this information to find trends and patterns in gestational age for different
groups of people. The model becomes a useful way to make rules based on facts. These helps
boost health for moms and kids, while helping public health projects forward.
Value Proposition:

Adding the ML model to these business steps creates several important benefits. First, it helps
in healthcare by making prenatal care plans more accurate. This accuracy turns into better
health results for both moms and babies. Next, the model helps hospitals use their resources
well. This makes sure that things like medical care and services for pregnant women are given
out in a smart way to meet different amounts of need. This efficiency helps make healthcare
services better. Being able to aim public health efforts based on what the model thinks will
happen is another important benefit. Health groups can tell where they need to help most,
dealing with special mother health problems in various places. Adding the prediction of baby's
age at birth into insurance risk models gives insurers a better knowledge about healthcare costs
linked with it. This makes decisions about insurance costs and coverage better for both the
companies that sell it and people who buy it. Finally, the model's knowledge backs up policy
making based on facts for maternal and child health. By looking at how gestational age changes
in different groups of people, scientists and those who make laws can create rules based on
hard data. This study-backed method helps make good rules to better mom and baby health. It
lines up bigger public goals for getting people healthy too. In short, using an ML model to

7
change these processes makes decisions based on data better and also improves efficiency and
health outcomes for moms.
Task:02
Part 1.
1. Accessing external sources of data including data lakes.

BigQuery ML is a way to do machine learning on Google's BigQuery platform. It helps users create and
use models easily in the SQL system of BigQuery. The first studies talk about the big change that comes
with BigQuery ML. They stress how good it is to combine machine learning directly into where we do
data analysis work(Kathiravelu and Sharma, 2017). Studying in the wider field of computer learning
shows how important data is for training models and why it's good to use different sources of
information(Derakhshannia et al., 2020). Articles by experts talk about the problems that come with
adding outside data to machine learning processes. They also discuss how it affects model accuracy and
ability to work on different tasks. As companies start using data lakes to store lots of different types of
information, the writing shows that there's a need for tools and systems. These should be able to easily
get at and study all this stored info. People talk about how BigQuery ML fits into data lake designs. It can
handle different file types and lets people look at outside information without having to move the data
around (Hagstroem et al., 2017). Examples of real-world businesses using BigQuery ML to get data from
outside places are found in case studies and industry news. These studies show real uses, problems
faced, and the good things achieved. Magazines for industry may give tips on how to use BigQuery ML
with data lakes. This can help in making decisions and predictions more effectively. Looking at how
BigQuery ML features have changed helps us understand what the platform has done to meet needs
over time. Writing may show big changes or additions to get data from the outside, showing that Google
Cloud is dedicated to having top-notch machine learning skills (Sarramia et al., 2022). BigQuery ML is
often checked in research papers and business reports for how well it works on access to data or
connecting with outside sources. Addressing challenges related to data formats, permissions, and
connectivity is crucial for a comprehensive understanding of the platform's capabilities(Hai, 2020).

2. Current Features and Capabilities:

BigQuery ML offers powerful tools to get data from other sources. This includes the EXTERNAL_QUERY
feature for getting information from outside. This allows people to use SQL commands on data stored in
other systems without transferring the info to BigQuery tables. Combining outside data with machine

8
learning helps make it easier to get information ready and improves how good models are made.
BigQuery ML is a great option because it can manage various types of information. The system can
handle common types like CSV, JSON, Avro and Parquet (Ciaccia, Martinenghi and Torlone in 2022). This
ability to change is extremely useful when dealing with data lakes, where different kinds of information
can be found. People can quickly work with various types of data from other sources, without having to
change hard formats a lot. BigQuery ML provides excellent results when fetching data from other places.
(Halevy et al., 2016) The picture shows that this system is fast for searching and uses its resources wisely.
These measures make it simpler and faster for users to use outside data sources. This ensures that we
can create machine learning models without wasting too much time or extra hours (Halevy et al., 2016).

3. Challenges and Considerations:

A big issue with getting information from other places is how slow the internet can be. When getting
data from distant lakes, time can affect how fast machine learning processes work. BigQuery ML makes
searching fast. You should check your network setup to shorten wait times and speed up data access. It's
very important to make sure the right people can get data when using BigQuery ML with outside
sources. (Grossman, 2019) People should be allowed to get info from other sources, and businesses
need good safety steps so they don't lose their secrets or private things. It's crucial to understand and
manage access permissions. This helps protect data from being used wrongly and follows rules for how
secure it should be (Wieder and Nolte, 2022). Getting good quality and same information is difficult,
especially when using data from other sources that can change shape or meaning. People using
BigQuery ML must put in place processes for checking data and create ways to deal with changes in the
outside information setup(Halevy et al., 2016). Keeping data good makes sure that machine learning
models trained on outside sources are trustworthy. BigQuery ML is a powerful tool, but it's important to
watch costs(Zone et al., no date). This especially matters when working with lots of data from other
sources outside BigQuery ML(Errami et al., 2023). Organizations must manage their data saving and
search costs. They need to use wisely so they can find a balance between performance and cost
savings. Choosing if you want to move data from outside sources into BigQuery tables or just look at the
data where it is, needs careful thinking. Even though the EXTERNAL_QUERY feature reduces data
moving, there are times where it could be good for speeding things up. People need to look at their own
reasons for using something and decide the best way(Grossman, 2019).

Part 2.
1. Model Explainability Using SQL in BigQuery ML: A Comprehensive Appraisal

9
Understanding how models work is very important in machine learning. It helps us know why predictions
are made the way they are. In BigQuery ML by Google, a tool for machine learning with BigQuery, they
help you understand model choices better using SQL features. This check looks closely at how much help
BigQuery ML gives for explaining models(Trenčeva et al., 2022). It studies the tools and methods that are
used, along with what needs to be thought about when doing so. BigQuery ML provides various features
to make models easier to understand. The main part is the ML.EXPLAIN function, a way that uses SQL to
create reasons for the trained machine learning model. This task lets people see how the model works.
They can learn what is important, why predictions happen and overall behavior of the system from this
task(Agrawal et al., 2019).

2. Utilizing the ML.EXPLAIN Function:

BigQuery ML's EXPLAIN function is very important for users that want to learn how their machine
learning models work. This job lets you get clear answers about single guesses, reasons for them and
how important the features are. Using SQL commands, people can easily connect the ML.EXPLAIN
function to their analysis process with BigQuery ML(Agrawal et al., 2019). The way you write this
command usually needs you to tell which model will be explained and what data features need an
explanation or reason behind them. People can use this tool's results in their SQL questions. This helps
them fully understand what makes a model guess right or wrong. This way of using SQL with BigQuery
ML not only works well but also makes it easy to include explanations about models into wider
studies(Jain and Kumar, 2023).

3. Feature Importance Analysis:

A big part of making models easy to understand is looking at how important certain features are, and
BigQuery ML does this very well. It gives users what they need to check the importance of these things
clearly. The ML. EXPLAIN function helps you understand how important each part is in the model's
choice process(Lakshmanan, 2022). These importance scores measure how important each feature is for
making predictions. This helps users to pay more attention and focus on key factors in their work. By
analyzing the importance of features, we can check and understand our model better. People can check
if the model matches their understanding and expectations about a subject. This feature is very useful in
situations where being clear and easy to understand are most important, like those found in controlled
businesses or places where the user's trust matters a lot(Pfaff et al., 2023).

4. Querying Explanatory Information

10
One good thing about how BigQuery ML explains its models is because it uses SQL to search for
information that helps understand them. After the ML. EXPLAIN function is used, people can make SQL
questions to get special info they want. This ability lets people closely study the reasons, letting them
concentrate on certain predictions or parts of data(Trad, 2023). Asking for information gives an easy and
fun way to look at how the model works. People can keep adjusting their questions to get more details,
fix guesses from the model or check how much changes affect it. This ability to ask smart questions helps
the user better understand and make sense of what a model is saying(Ligęza et al., 2022).

6. Visualizing Explanatory Data:

While SQL queries are the key for understanding models in BigQuery ML, this system also helps to
picture data that explains it. People can use different ways to show pictures andcreate easy-to-
understand info from the ML. EXPLAIN function. This picture part adds another side to the study, making
it easier for many people to understand. Showing important data can be done through plots of feature
importance, decision lines or other pictures that explain how the model makes choices. This not only
helps talk to others but also improves the total understanding of difficult computer learning models,
making them easier for people involved to use.

8. Industry Adoption and Best Practices:

Many different industries are starting to use explainable features in BigQuery ML. Groups understand
how important it is to have clear and easy-to-understand computer learning models. This is especially
true in areas where rules must be followed or trust from others matters a lot. As decision-making based
on data becomes more normal, there is a growing acceptance of tools like ML.EXPLAIN in BigQuery ML
because they make models easier to understand and check(Huang et al., 2023). In areas where rules are
needed such as money control, health care or insurance it's often required by law for results from these
decisions can be explained clearly so everyone knows what happened during the process. BigQuery ML
helps make models easy to understand, which is important for rules(Matos et al., 2023). This lets
companies use simple-to-understand models while following the right standards in their work
processes. Good practices in the industry highlight adding model explanations regularly into how we
check models. People who depend on it like data experts, thinkers and business leaders should use the
explainable features of BigQuery ML. This will help them check if model guesses are right or
wrong. Companies should set up simple ways to share information about outcomes from model use,
especially when choices have big effects(Jindal and Leeka, 2022).

11
9. Continuous Improvement and Updates:

The area of helping people understand how models work is always changing and getting better as more
research happens. As part of the bigger machine learning world, BigQuery ML is expected to keep getting
better in how it explains its models. These improvements might involve making current features better,
introducing new ways to understand things clearly and changes for specific issues faced by different
kinds of learning in machines(Russell et al., 2022). Google Cloud shows it's working hard on making
machine learning better. This is part of BigQuery ML too, which always gets new features and
improvements over time. People can expect that Google Cloud will keep doing good work on making
models easy to understand. They'll add the latest methods into BigQuery ML so they can face new issues
and what businesses need in future times. People's opinions are important in making BigQuery ML
better. This includes understanding how the model works. When companies give their opinions on how
good and useful the service is, Google Cloud will probably use these thoughts to make it better(Kashyap,
2023).

10. Conclusion:
Big Query ML shows how it helps machine learning a lot in a SQL-based world. The good work
of making a machine learning model that guesses gestational age shows how well the platform
works with many types of data and lets it easily fit into what we're already using to check
facts. Following the best methods, like making separate training and testing data sets shows a
promise to strong model building and checking processes. The platform's ability to get data
from outside sources like data lakes is a main thing it does well. School papers and reports show
that companies gain from Big Query ML. This lets them look at data straight from other sources
without moving the data. This not only makes the model development process faster but also
matches with how data storage and processing are changing. A big check on how clear models
is in Big Query ML shows us a set of tools that makes them more understandable and easier to
see. The ML.EXPLAIN function, powered by SQL and used with feature importance analysis,
querying information that helps explain choices made by machine learning models, gives users
the power to understand and share those decisions without confusion. This ability is very
important for getting user trust, meeting laws needed and encouraging use in businesses.
The report talks about problems and thoughts linked to getting information from outside. It
underlines that issues such as how quickly a network works, who is allowed to use it, the quality

12
of data and cost factors are tricky things. This important look shows why we need a fair plan in
using Big Query ML for gathering data from other places. It must work fast and be good value
too. s more people use Big Query ML, it shows that they see value in this tool. This is especially
important for controlled fields where understanding what a model does or doesn't explain well
matters a lot to them. The regular changes and updates in Big Query ML, based on user opinion
and new learning research progress, make it a forever-improving tool. Adding a measure to the
result, real-life examples show that companies using Big Query ML for machine learning files get
faster results in their questions. They also become easier to expand and use resources well. The
numbers show that the system quickly deals with big data from outside. This helps to make it
easy and fast for users.

10. References:
Agrawal, A. et al. (2019) ‘Cloudy with high chance of DBMS: A 10-year prediction for Enterprise-Grade
ML’. arXiv. Available at: http://arxiv.org/abs/1909.00084 (Accessed: 31 December 2023).

Ciaccia, P., Martinenghi, D. and Torlone, R. (2022) ‘Conceptual Constraints for Data Quality in Data Lakes’,
in CEUR WORKSHOP PROCEEDINGS. CEUR-WS, pp. 111–122. Available at: https://ceur-ws.org/Vol-
3340/paper34.pdf (Accessed: 31 December 2023).

Errami, S.A. et al. (2023) ‘Spatial big data architecture: From Data Warehouses and Data Lakes to the
LakeHouse’, Journal of Parallel and Distributed Computing, 176, pp. 70–79.

Grossman, R.L. (2019) ‘Data lakes, clouds, and commons: A review of platforms for analyzing and sharing
genomic data’, Trends in Genetics, 35(3), pp. 223–234.

Hai, R. (2020) Data integration and metadata management in data lakes. PhD Thesis. Dissertation,
RWTH Aachen University, 2020. Available at:
https://scholar.archive.org/work/47amdbt3pncpdm66pc3glp36um/access/wayback/https://
publications.rwth-aachen.de/record/795304/files/795304.pdf (Accessed: 31 December 2023).

Halevy, A.Y. et al. (2016) ‘Managing Google’s data lake: an overview of the Goods system.’, IEEE Data Eng.
Bull., 39(3), pp. 5–14.

13
Huang, Z. et al. (2023) ‘JoinBoost: Grow Trees Over Normalized Data Using Only SQL’. arXiv. Available at:
http://arxiv.org/abs/2307.00422 (Accessed: 31 December 2023).

Jain, S. and Kumar, P. (2023) ‘Cost Effective Generic Machine Learning Operation: A Case Study’, in 2023
International Conference on Data Science and Network Security (ICDSNS). IEEE, pp. 1–6. Available at:
https://ieeexplore.ieee.org/abstract/document/10245408/ (Accessed: 31 December 2023).

Jindal, A. and Leeka, J. (2022) ‘Query Optimizer as a Service: An Idea Whose Time Has Come!’, ACM
SIGMOD Record, 51(3), pp. 49–55. Available at: https://doi.org/10.1145/3572751.3572767.

Kashyap, R. (2023) ‘Machine Learning in Google Cloud Big Query using SQL’, SSRG International Journal of
Computer Science and Engineering, 10(5), pp. 17–25.

Lakshmanan, V. (2022) Data Science on the Google Cloud Platform. O’Reilly Media, Inc. Available at:
https://books.google.com/books?
hl=en&lr=&id=F7tmEAAAQBAJ&oi=fnd&pg=PP1&dq=1.%09Model+Explainability+Using+SQL+in+BigQuer
y+ML&ots=Sz5bS7ynZU&sig=iG-fL-slLZTNV2bOE7p7bU7nYbw (Accessed: 31 December 2023).

Ligęza, A. et al. (2022) ‘Evaluation of Selected Artificial Intelligence Technologies for Innovative Business
Intelligence Applications’, in L. Borzemski, H. Selvaraj, and J. Świątek (eds) Advances in Systems
Engineering. Cham: Springer International Publishing (Lecture Notes in Networks and Systems), pp. 111–
126. Available at: https://doi.org/10.1007/978-3-030-92604-5_11.

Matos, J. et al. (2023) ‘Shining Light on Dark Skin: Pulse Oximetry Correction Models’, in 2023 IEEE 7th
Portuguese Meeting on Bioengineering (ENBENG). IEEE, pp. 211–214. Available at:
https://ieeexplore.ieee.org/abstract/document/10175316/ (Accessed: 31 December 2023).

Pfaff, E.R. et al. (2023) ‘De-black-boxing health AI: demonstrating reproducible machine learning
computable phenotypes using the N3C-RECOVER Long COVID model in the All of Us data repository’,
Journal of the American Medical Informatics Association, 30(7), pp. 1305–1312.

Russell, M. et al. (2022) ‘Machine learning for surgical risk assessment decision systems’, in 2022
International Joint Conference on Neural Networks (IJCNN). IEEE, pp. 1–8. Available at:
https://ieeexplore.ieee.org/abstract/document/9892752/ (Accessed: 31 December 2023).

14
Trad, A.T. (2023) ‘Enterprise Transformation Projects/Cloud Transformation Concept: The Compute
System (CTC-CS)’, in Handbook of Research on Advancements in AI and IoT Convergence Technologies. IGI
Global, pp. 145–177. Available at: https://www.igi-global.com/chapter/enterprise-transformation-
projectscloud-transformation-concept/330064 (Accessed: 31 December 2023).

Trenčeva, Ž. et al. (2022) ‘198. BIGQUERY FOR BIG DATA ANALYSIS (BIGQUERY за анализа на големи
податоци)’, Journal of Electrical Engineering and Information Technologies, pp. 77–85.

Wieder, P. and Nolte, H. (2022) ‘Toward data lakes as central building blocks for data management and
analysis’, Frontiers in big Data, 5, p. 945720.

Zone, B.-A.T.D.P. et al. (no date) ‘Demand-Driven Data Provisioning in Data Lakes’. Available at:
https://www.ipvs.uni-stuttgart.de/departments/as/publications/stachch/iiwas_21_barents.pdf
(Accessed: 31 December 2023).

15

You might also like