Deep Neural Networks for Supervised Learning: Classification
In Chapter 3 , we explored a DL use case for regression. We explored the entire problem-solving approach with a business-forward strategy. We leveraged all our learning from Chapters 1 and 2 in foundational DL and the Keras framework to develop DNNs fo
- PDF / 453,699 Bytes
- 35 Pages / 439.37 x 666.142 pts Page_size
- 19 Downloads / 292 Views
Deep Neural Networks for Supervised Learning: Classification In Chapter 3, we explored a DL use case for regression. We explored the entire problem-solving approach with a business-forward strategy. We leveraged all our learning from Chapters 1 and 2 in foundational DL and the Keras framework to develop DNNs for a regression use case. In this chapter, we will take our learning one step further and design a network for a classification use case. The approach overall remains the same, but there are a few nuances we need to keep in mind while solving a classification use case. Moreover, we will take our learning in this chapter one step ahead with extensive DNN architectures. Let’s get started.
Getting Started Similar to Chapter 3, we will consider Kaggle for our use case’s data source. From the available options, we will use the dataset provided for the “Red Hat Business Value” competition. This competition was hosted on Kaggle a few years back, and the dataset is a really good business use case for our study. The archived competition is available at www.kaggle.com/c/ predicting-red-hat-business-value. Just as in the previous use case, © Jojo Moolayil 2019 J. Moolayil, Learn Keras for Deep Neural Networks, https://doi.org/10.1007/978-1-4842-4240-7_4
101
Chapter 4
Deep Neural Networks for Supervised Learning: Classification
we need to read and accept the competition rules before downloading the dataset for our experiments. Once you have accepted the competition rules, you can download the dataset from the “Data” tab or www.kaggle. com/c/predicting-red-hat-business-value/data. The data will be downloaded as a zip file. After unzipping, you will have four different datasets. We will need only two of them: act_train.csv and people.csv. You can copy these two datasets and keep them in a new folder for the current chapter’s experiments. We will use the same environment for the use case, but before we begin, let’s have a look at the problem statement and define the SCQ and the solution approach, just like we did in Chapter 3.
Problem Statement The high-level problem statement is mentioned in the competition’s description page. It highlights the problem that deals with predicting high- value customers for their business based on the operational interaction data and thereby helping the company effectively prioritize resources to generate more business and serve its customers better. Let’s have a look at the problem statement from a more business- centric view. We will start by understanding the customer better. The organization is an American multinational software company that provides open source software products to the enterprise community. Their primary product is Red Hat Enterprise Linux, the most popular distribution of Linux OS, used by various large enterprises. In its services, it helps organizations align their IT strategies by providing enterprise-grade solutions through an open business model and an affordable, predictable subscription model. These subscriptions from large enterprise customers create a
Data Loading...