Build a Deep Learning Python Model that forecasts CO2 levels
- Description
- Curriculum
- FAQ
- Reviews
What is the course about:
The course teachesm from scratch, the development of a Deep Learning model that forecasts the levels of CO2 emissions in India, UK, USA, European Union, China, South Africa; the code is in Python and is fully downloadable.
At the end of the course you will confidently feel able to build such a model from scratch, a knowledge valuable in our Climate Change and Data Science era.
Exclusive
We are new on Udemy and to Celebrate this we are offering this online course only for $7 , along with its e-book, to the next 50 students who message us at www datasciencedatabase com.
Who:
Dr. Spyridon is a Senior Research Scientist and is leading a Team of Researchers working in the Data Science Database project, which includes online courses, each of which teaches how to build from scratch to completion one specific Data Science model – as opposed to regular online courses that teach generic knowledge/skills without any specific deliverable in the end, causing more students to have low confidence feeling they spent so much time without actually learning anything of value.
Dr. Spyridon has a PhD from Imperial College London in Mathematical Optimization and Data Science.
You can pre-enroll in the Data Science Database project and receive notifications at www datasciencedatabase com
-
1
Summary - bullet pointsIf you have questions, you can share them with us at
www datasciencedatabase com. -
2
Why do we apply Deep Learning to CO2 emissions?Understanding the goal of this course.
-
3
B O N U S #1: Perpetual Support
Your first B O N U S is Perpetual Support i.e. ask as many questions, any time, and receive answers from the Instructors within hours. No need to waste time google searching!
-
4
B O N US #2 : Perpetual Maintenance
What it means to have "perpetual maintenance": The code of the code will be updated every time a new version of Python comes out. And the videos will be recorded again to reflect the updated code! For life!
-
5
B O N U S #3: Your questions are replied by email, by recording a video or skype
What it means to go one step further in our support !
-
6
B O N U S #4: Free ebook summarising this course!
We have an ebook for this course!
-
7
B O N U S #5 : Coupons available !
We understand you would like better prices!
-
8
Description and Implementation of Data Preprocessing
We explore the WorldBanks' Databank , and we download the dataset directly into our code. We then reformat the dataset and bring it to the correct form. We remove missing values, we correct the datatypes of the different columns, and we interpret the data.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
9
What does the pivot table command do?
-
10
Why is Data Preprocessing needed?
We look at the first step of building our model.
-
11
Defining the features and the features matrix
In the beginning of every Deep Learning model we need to create the 'features' and the 'features matrix'.
We also explain what 'bias' means.Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
12
Polynomial Features Transformer
Why use it.
-
13
Splitting the original dataset into its Training and Test components
In this video we define the target variables.
Then we split the original dataset at 80%-20% ratio. So the training part will be 80% of the original dataset. The remaining will be the test set.
We then split the Features matrix as well as the target variables in two parts: the training and the test component.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
14
The logic of the split.
Why is the split as it is shown in the videos.
-
15
Scaling the features and the target variables
We scale the features matrix and the target variables. This means that we transform their values and bring them to a form needed by the Deep Neural Networks.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
16
The use of scaling.
Which algorithms generally need scaling.
-
17
Defining and compiling the Deep Neural Network Models
We define the deep neural network models and then we compile them. These are necessary steps prior to fitting / training the models, which will be described in the next lecture.
We also explore the structure of a Deep Neural Network, explaining its individual components.
We also go through the concept of the activation function.Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
18
Fitting the Deep Neural Network models
We fit the Deep Neural Network models to the training datasets.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
19
Drawing the Deep Neural Network and clarifying the Activation Function
In this video we explore the structure of the Deep Neural Networks and its individual components. This requires the installation of special packages in Python, which we do step by step.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
20
Showing how to draw it in more detail
In this video we mention some specific topics related to drawing the Deep Neural Network models.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
21
The fitting process.
Understanding what the fitting process is all about.
-
22
Generating the training and test set predictions
We generate the training set and the test set predictions. These are the outputs of the Deep Learning models for when the inputs are the training and the test sets.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
23
Predictions and Forecasts
Understanding the difference
-
24
Generating the test set Errors
We generate the errors on the test set predictions. These are defined as the difference between the outputs of the model on the test set, and the actual test-set values.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
25
Generating the training set Errors
We generate the errors on the training set predictions. These are defined as the difference between the outputs of the model on the training set, and the actual training-set values.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
26
The essence of finding the errors.
Test and Training errors
-
27
Conducting overfitting analysis
We conduct overfitting analysis, which means we check the models in terms of whether they overfit or not.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com.If you have questions, you can share them with us at
www datasciencedatabase com. -
28
Hyperparameter tuning versus Sensitivity Analysis
We go through the differences between hyperparameter tuning and Sensitivity Analysis. These are fundamental concepts in Machine Learning.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
29
Sensitivity analysis : Test set MAPE based on hyperparameters
We conduct sensitivity analysis in order to discover more about the behaviour of the models in terms of the hyperparameters used.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
30
The Naive Model Benchmark test
We conduct the naive-model benchmark test, which allows us to evaluate whether or not the test-set errors that we have generated can be considered unacceptably large that would disqualify the models from being used to generate the forecasts.
Make sure you activate the subtitles, which are carefully written; they are not auto-generated.
You can also read the transcript and download the code.
If you have questions, you can share them with us at
www datasciencedatabase com. -
31
What is actually overfitting?
Understanding the concept of overfitting.
