AI Portfolio Project | I built a MACHINE LEARNING MODEL using AI in 10 MINUTES

Mo Chen
22 Jan 202409:58

Summary

TLDRThe video script showcases the power of AI tools like pcan in simplifying the process of building powerful machine learning models, even without coding experience. Through an interactive conversation with an AI assistant, the host walks through the steps of creating a revenue forecasting model at the customer level, from data upload and analysis to model training and evaluation. The process is remarkably streamlined, demonstrating how AI has lowered the barriers to learning and utilizing advanced technologies. However, the host emphasizes that while AI tools have made model building more accessible, coding skills and technical expertise remain invaluable for fine-tuning and optimizing models.

Takeaways

  • ๐Ÿ“ˆ AI-powered tools like pcan have significantly lowered the barrier to building machine learning models, allowing even beginners to create powerful models with minimal coding.
  • ๐Ÿ”‘ The script illustrates a step-by-step process of building a revenue forecasting model at the customer level using pcan, without any coding involved.
  • ๐Ÿง  pcan's AI assistant guides the user through a series of questions to understand the desired model, data requirements, and prediction targets.
  • ๐Ÿ“Š pcan automatically generates SQL queries, handles data processing, and constructs the core dataset (coret) required for model training.
  • ๐Ÿ›  Users can choose to train their model for faster results or better performance, with pcan's powerful engine enabling highly accurate predictions.
  • ๐Ÿ“ˆ The platform provides comprehensive model evaluation metrics, visualizations, and interpretability features to assess model performance and importance of input features.
  • ๐Ÿค– While AI tools simplify model building, coding skills and technical expertise are still valuable for fine-tuning models and leveraging advanced capabilities.
  • ๐Ÿงฎ Strong mathematical and statistical knowledge remains essential for developing robust machine learning models.
  • โšก The rapid advancement of AI tools has significantly accelerated the process of building complex models, making it more accessible to a wider audience.
  • ๐ŸŒŸ pcan demonstrates the potential of AI-powered tools to democratize machine learning and empower users of all skill levels to create impactful models efficiently.

Q & A

  • What is the main topic of the video script?

    -The script discusses how to build a powerful machine learning model with no actual coding involved, using an AI assistant tool called Pcan.

  • What kind of machine learning model is being built in the example?

    -The example demonstrates building a revenue forecasting model at the customer level, predicting the revenue for each customer on a monthly basis.

  • What data set is used for building the machine learning model?

    -The data set used is a cleaned-up Kaggle competition file, the link for which is provided in the video description.

  • How does the Pcan AI assistant guide the model building process?

    -The Pcan AI assistant asks guiding questions to understand the requirements, recommends data column mappings, generates SQL queries, and constructs the core data set for training the model.

  • What is the significance of the 'attribute' section in the generated notebook?

    -The 'attribute' section contains the data that the model will use to identify and find patterns to produce the correct target values (in this case, revenue predictions).

  • What are the two training options provided by Pcan, and what is the difference between them?

    -The two training options are 'fastest' (which takes 10-30 minutes) and 'production quality' (which takes longer but provides better performance). The script mentions that the production quality option took about an hour to train the model.

  • How does the script evaluate the performance of the trained machine learning model?

    -The script evaluates the model's performance using various metrics and visualizations provided by Pcan, such as model precision, column importance, and model output (predictions for each user and month).

  • What is the author's view on the importance of coding skills in the age of AI-assisted model building?

    -The author acknowledges that while AI tools can simplify model building, coding and technical skills are still very important for fine-tuning models and understanding the underlying math and statistics.

  • What is the main advantage of using AI tools like Pcan for building machine learning models, according to the script?

    -The main advantage highlighted is that AI tools like Pcan lower the barrier to learning and building machine learning models, allowing users to create powerful models quickly, even with limited coding experience.

  • What is the author's overall impression of the Pcan tool's capabilities?

    -The author seems impressed by Pcan's capabilities, describing the process as "AI magic" and expressing amazement at how AI can simplify complex tasks like building strong machine learning models through conversational interactions.

Outlines

00:00

๐Ÿค– Revolutionizing Machine Learning with AI Assistants

The narrator is amazed by the ability to build powerful machine learning models simply by conversing with an AI assistant, a stark contrast from the difficulties faced earlier in their career. They highlight how AI tools and services have lowered the barrier to learning, allowing for quicker knowledge acquisition. In this video, the narrator aims to demonstrate the simplest way to build a robust machine learning model without coding, using the AI assistant Pcan, which has sponsored the video.

05:01

๐Ÿง  Step-by-Step Guide to Building a Revenue Forecasting Model

The narrator walks through the process of building a revenue forecasting model using Pcan's AI assistant. They answer guided questions to specify the prediction task, target value, and time horizon. Pcan generates a predictive analytics question and assists in data connection and schema mapping. The narrator explains the concept of the 'coret' (training data) and attributes (data used for pattern recognition). Pcan constructs SQL queries, handles data sampling, and provides options for model training. The narrator evaluates the model's performance using metrics like precision, column importance, and model output.

Mindmap

Keywords

๐Ÿ’กMachine Learning

Machine learning is a field of artificial intelligence that enables computer systems to learn and improve from experience without being explicitly programmed. In the context of this video, machine learning refers to the process of building models that can make predictions or decisions based on data. The video demonstrates how an AI assistant can guide users through the process of building a powerful machine learning model for revenue forecasting without the need for extensive coding.

๐Ÿ’กAI Assistant

An AI assistant is an artificial intelligence-powered software program that can understand and respond to human queries, commands, or inputs through natural language interactions. In this video, the AI assistant provided by the platform Pcan is used to guide the user through the process of building a machine learning model by asking guiding questions, analyzing data, generating SQL queries, and training the model. The AI assistant simplifies the process of creating machine learning models, making it accessible to users with varying levels of technical expertise.

๐Ÿ’กPredictive Analytics

Predictive analytics is a branch of advanced analytics that uses statistical models, machine learning techniques, and data mining to make predictions about future outcomes or events. The video focuses on building a predictive model for revenue forecasting at the customer level, which means predicting the expected revenue for each individual customer in the future. The AI assistant guides the user through defining the predictive question, selecting the target value (revenue), and specifying the time horizon and frequency of predictions.

๐Ÿ’กData Set

A data set is a collection of data, often organized in a structured format such as a spreadsheet or database table. In the video, the user uploads a CSV file containing customer data, which the AI assistant analyzes and utilizes to build the machine learning model. The availability of relevant and high-quality data is crucial for training accurate and reliable predictive models.

๐Ÿ’กNotebook

A notebook in the context of machine learning is an interactive environment that allows users to write, run, and document code, as well as visualize results. In the video, the AI assistant generates a predictive notebook that contains the SQL queries, data transformations, and modeling steps required to build the revenue forecasting model. The notebook serves as a central workspace for the user to explore and understand the model's logic and construction.

๐Ÿ’กTarget Value

The target value, also known as the dependent variable or the variable to be predicted, is the output or outcome that a machine learning model aims to forecast or classify. In the case of this video, the target value is the revenue for each customer, which the model tries to predict based on the available attributes or features in the data set.

๐Ÿ’กAttributes

Attributes, also known as features or independent variables, are the input data or characteristics used by a machine learning model to make predictions or classifications. In the video, the AI assistant identifies the relevant attributes from the data set, such as the customer ID, transaction amount, and event time, which the model can use to find patterns and relationships that will help predict the target value (revenue).

๐Ÿ’กModel Training

Model training is the process of feeding a machine learning algorithm with data to learn the patterns and relationships that can be used to make accurate predictions or classifications. In the video, the user has the option to choose between a faster training process or a more computationally intensive, production-quality training process that may yield better performance. The AI assistant automates the training process, taking care of data validations and model optimization.

๐Ÿ’กModel Evaluation

Model evaluation is the process of assessing the performance and accuracy of a trained machine learning model using various metrics and visualizations. The video showcases the model evaluation tab, which provides insights into the model's precision, column importance (the most influential features), and individual predictions for each customer and month. Evaluating the model's performance is crucial for determining its reliability and identifying areas for improvement.

๐Ÿ’กNo-Code

No-code refers to the ability to build applications or models without traditional coding or programming. In the context of the video, the AI assistant provided by Pcan allows users to build a powerful machine learning model for revenue forecasting without the need for writing any code. The guided questions, data analysis, query generation, and model training are all handled by the AI assistant, making machine learning more accessible to users with varying technical backgrounds.

Highlights

I can, build a powerful machine learning model, nowadays just by chatting to an AI, assistant

With AI, tools and services on the rise the, barrier to learning anything really has, been lowered or at the very least you, can learn whatever you want quicker if, you know what tools to use and how to, use them in the most efficient way

I'll show you the simplest, way I know to build a strong machine, learning model with no actual coding, involved

The way the, assistant works is that it asked me some, guiding questions to help me build my, machine learning model

Pcan generated the Predictive Analytics, question for me all I had to do was to, confirm that everything still looked, good and I moved on to connecting to my, data

Pcan is um capable of, processing lots and lots of information, as it uses data bricks so you should, feel confident about loading massive, data sets

The AI assistant, looked at the schema of the file and, recommended the column mappings of the, data set based on the schema

Pcan, did all the heavy lifting for me in the, background so now all I had to do was to, click on generate notebook and then go, to it to see what's actually in there

Think of the predictive notebook that I, just generated as the brains or the, control room of the entire model, building process

Pcan's AI, assistant constructed the core set for, me by using the answers I provided to, the guided questions

An attribute is the, data that my model will use to identify, and find the patterns that will then, tell my my model what the predictions, will be

Pcan also has, the option to add more attributes if I, wanted to so I could have easily added, more by adding more data

I hit run all which ran all of, my SQL queries, then I trained my model

Pcan quickly rans some data, valid ations to check that my data was, actually fit for building a predictive, model and that was it I sat back and, waited for about an hour and I had my, Model results with some cool metrics and, visualizations

It is still, mindboggling to me that it allows me to, do such difficult tasks like building a, strong machine learning model just by, interacting with an AI chat assistant

Transcripts

00:00

I truly think it's insane that I can

00:02

build a powerful machine learning model

00:04

nowadays just by chatting to an AI

00:06

assistant I remember when I first

00:08

started my career doing Predictive

00:10

Analytics was hard really hard with AI

00:14

tools and services on the rise the

00:16

barrier to learning anything really has

00:18

been lowered or at the very least you

00:20

can learn whatever you want quicker if

00:22

you know what tools to use and how to

00:24

use them in the most efficient way in

00:27

today's video I'll show you the simplest

00:30

way I know to build a strong machine

00:32

learning model with no actual coding

00:34

involved I used pcan who have kindly

00:37

also sponsored this video but as always

00:40

all opinions thoughts and Reflections on

00:42

the tool will of course be absolutely of

00:45

my own whether you're a machine learning

00:47

newbie or have some or maybe lots and

00:50

lots of experience training your own

00:51

models I'd encourage you to stick around

00:54

as you'll get to see how simple and

00:56

effective it can be to utilize the

00:58

latest Technologies to build build your

01:00

own machine learning models regardless

01:02

of your experience and I say regardless

01:05

of experience because even though you

01:07

can build precise models with zero

01:10

coding involved you can of course adjust

01:12

your code within pcan I'll show you

01:14

later how so I'll walk you through how I

01:17

built a revenue forecasting model but

01:19

you could obviously predict other things

01:21

like LTV customer turn or winback

01:24

marketing mix or how to up and cross

01:26

sell you can try and build your own

01:28

model the same way by the way ass

01:30

signing up to pcan is completely free

01:32

I'll put the link in the description

01:34

below along with the link to the exact

01:37

data set that I used so after I logged

01:40

in straight away I got to interact with

01:42

the pcan AI assistant the way the

01:44

assistant works is that it asked me some

01:47

guiding questions to help me build my

01:49

machine learning model so I answered the

01:52

first question and told the AI assistant

01:55

that I would like a revenue forecast

01:56

model at the customer level this simply

01:59

means that I want to predict what the

02:01

revenue is for each and every customer

02:04

the assistant is pretty smart as it

02:06

recommended the subject of my prediction

02:09

which was the customer of course and my

02:11

target value which was the revenue and

02:14

all I had to do was to confirm that all

02:17

was good so far then I got another

02:19

guiding question asking me how far into

02:21

the future I'd like to predict the

02:23

revenue for each customer you can choose

02:26

whatever you want really next week month

02:28

quarter or even year but I went with

02:31

probably the most popular future time

02:33

Horizon a month year then the tool asked

02:36

me on what recurring basis I'd like to

02:38

make these Revenue predictions again I

02:41

went with something simplistic here so I

02:43

said that I'd like the frequency of the

02:45

predictions to be monthly I also had the

02:48

options to trigger my predictions based

02:50

on specific events for example when a

02:52

customer makes a purchase but for this

02:54

exercise for this model I chose to go

02:57

with predicting on a monthly basis then

03:00

pcan generated the Predictive Analytics

03:03

question for me all I had to do was to

03:05

confirm that everything still looked

03:07

good and I moved on to connecting to my

03:10

data I uploaded my CSV file which is

03:12

actually a kaggle competition file that

03:14

was cleaned up a bit and as I said

03:16

before the link is in the description

03:18

below so you can go ahead and download

03:21

it and build your own machine learning

03:23

model now of course pcon has a bunch of

03:26

connectors when it comes to connecting

03:28

to your data sets so if you have

03:30

something sitting on SQL servers or with

03:33

popular Cloud providers that's not a

03:36

problem at all pcan is um capable of

03:39

processing lots and lots of information

03:41

as it uses data bricks so you should

03:43

feel confident about loading massive

03:46

data sets after the file upload was

03:48

complete I simply sent it to the chat

03:51

and let the tool do its work with

03:52

analyzing the data the AI assistant

03:55

looked at the schema of the file and

03:57

recommended the column mappings of the

03:59

data set based on the schema for each

04:01

column type I quickly saw what the data

04:03

type was I confirmed that everything

04:06

looked good and moved on this time the

04:09

tool actually figured it out that the

04:11

user ID column was the customer ID the

04:14

amount represented the revenue for each

04:16

transaction and the event time was the

04:19

date or timestamp that I needed again I

04:22

confirmed that everything looked good

04:25

and then I got a summary with my

04:27

predictive question the schema my Target

04:30

table Target value column Target table

04:32

date column and the target entity pcan

04:36

did all the heavy lifting for me in the

04:38

background so now all I had to do was to

04:41

click on generate notebook and then go

04:43

to it to see what's actually in there

04:46

think of the predictive notebook that I

04:47

just generated as the brains or the

04:50

control room of the entire model

04:52

building process I had my SQL queries in

04:56

here that the AI assistant built for me

04:58

you can also see that all of the steps

05:00

were clearly explained each query had a

05:03

name and was actually saved kind of like

05:06

a view so then the next query or sell

05:08

could utilize the data from the query

05:10

above for example the sampled customers

05:13

table could utilize the information from

05:15

the monthly sampling table so Pan's AI

05:18

assistant constructed the core set for

05:20

me by using the answers I provided to

05:23

the guided questions now what do I mean

05:26

by a coret it is the data that allowed

05:28

my model to to know what it needs to

05:30

learn and where it needs to learn from

05:33

so the coret was a final table that

05:35

consisted of all users sample dates and

05:38

the target value which in my case was

05:40

the revenue within 1 month after the

05:42

sample date now the last thing in my

05:45

notebook was the attribute section and

05:47

if you have no idea what this is don't

05:49

worry as I'm going to explain this in a

05:51

very simple way and attribute is the

05:54

data that my model will use to identify

05:57

and find the patterns that will then

05:59

tell my my model what the predictions

06:01

will be so when the model is training

06:03

it's learning the patterns to produce

06:06

the correct Target values long story

06:08

short use the training set to predict

06:11

your Target on an ongoing basis then use

06:14

your model in real life pcan also has

06:16

the option to add more attributes if I

06:18

wanted to so I could have easily added

06:21

more by adding more data whether it was

06:23

just me uploading some files or

06:25

connecting to one of the many databases

06:27

that are supported now of of course if I

06:30

added more data I could have just used

06:31

the AI assistant to replace my already

06:34

existing query so that it runs on my new

06:37

data set again no coding involved just

06:40

pure AI magic I wanted to keep my model

06:43

nice and light so I chose not to add any

06:45

more attributes I just went with the one

06:48

I already had my notebook had everything

06:51

I needed to train my model using the

06:53

data set I uploaded into the chat

06:55

earlier I hit run all which ran all of

06:58

my SQL queries

07:00

then I trained my model now I had two

07:03

options fastest which is by no surprise

07:06

fast as it only takes about 10 to 30

07:08

minutes to train a model and production

07:10

quality which provides better

07:12

performance but takes a little longer to

07:14

train I know it says several hours here

07:17

but my model was actually built in about

07:19

an hour or so the time it takes will of

07:22

course depend on how complex your

07:23

machine learning model is so once I hit

07:26

train model P can quickly rans some data

07:29

valid ations to check that my data was

07:30

actually fit for building a predictive

07:33

model and that was it I sat back and

07:35

waited for about an hour and I had my

07:37

Model results with some cool metrics and

07:40

visualizations that helped me to gauge

07:42

how good my model actually was I could

07:45

see on the model evaluation tab under

07:48

model performance that my model was very

07:51

precise meaning the model was very close

07:53

to the actual values the pan platform

07:56

has a really powerful engine behind it

07:58

hence the really good result results

07:59

clearly their data teams know what

08:02

they're doing now the model evaluation

08:04

tab has a bunch of other metrics as well

08:07

and I won't go through all of them but I

08:09

will highlight column importance which

08:11

was very useful to know in my model I

08:14

could see that the amount column was by

08:16

far the top contributing column to the

08:19

model predictions which makes sense

08:22

another helpful tab I used for

08:24

interpreting my Model results was the

08:26

model output tab I could quickly see

08:29

what the prediction was for each user

08:31

for each month I'm getting more and more

08:33

used to how much and how fast AI can do

08:35

things but at times it is still

08:37

mindboggling to me that it allows me to

08:40

do such difficult tasks like building a

08:42

strong machine learning model just by

08:44

interacting with an AI chat assistant

08:47

now does this mean that coding is

08:49

useless absolutely not the ability to

08:53

read write understand and interpret code

08:55

is still very very important in my

08:57

opinion just think of this who can

09:00

develop the better machine learning

09:01

model in a shorter time frame the person

09:04

who knows how to use the latest AI tools

09:06

but has no real technical expertise or

09:09

the person who also knows how to use the

09:12

latest AI tools and has extremely strong

09:15

technical skills surely the person with

09:17

the strong technical skills will be able

09:19

to fine-tune the model better and faster

09:21

using the same AI tools compared to the

09:24

person with little technical expertise

09:27

not to mention all of the underlying

09:29

math and statistical knowledge that is

09:31

of course also essential to building

09:33

good machine learning models anyway I'm

09:36

going to end this video here if you

09:38

enjoy content like this make sure to

09:41

check out some of my other videos right

09:43

here thank you so much for watching and

09:46

I shall see you in the next

09:51

[Music]

09:57

one