Applying machine learning algorithms in sales prediction
Loading...
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Kadir Has Üniversitesi
Open Access Color
OpenAIRE Downloads
OpenAIRE Views
Abstract
Makine öğrenimi bir çok endüstride üzerinde yoğun çalışmalar yapılan bir konu olmuştur, ve neyse ki şirketler kendi problemlerini çözebilecek çeşitli machine learning yaklaşımları hakkında günden güne daha fazla bilgi sahibi oluyorlar. Fakat, farklı makine öğreniminin modellerinden en iyi şekilde sonuç almak ve verimli sonuçlara ulaşabilmek için, modellerin uygulanış biçimlerini ve verinin doğasını iyi anlamak gerekir. Bu tez, belli bir tahmin görevi için, uygulanan farklı makine öğreniminin algoritmalarını ne kadar iyi sonuç verdiklerini araştırır. Bu amaçla tez, 4 faklı algoritma, bir istifleme topluluğu tekniği ve modeli geliştirmek için belirli bir özelllik seçme yaklaşımı sunar ve uygular. Farklı konfigürasyonlar uygulayarak sonuçlar birbiriyle test edilir. Bütün bu işlemler, gerekli veri önislemeleri ve özellik mühendisliği adımları tamamlandıktan sonra yapılır.
Machine learning has been a subject undergoing intense study across many different industries and fortunately, companies are becoming gradually more aware of the various machine learning approaches to solve their problems. However, in or- der to to fully harvest the potential of different machine learning models and to achieve efficient results, one needs to have a good understanding of the application of the models and of the nature of data. This thesis aims to investigate different approaches to obtain good results of the machine learning algorithms applied for a given prediction task. To this end the thesis proposes and implements a four different algorithms, a stacking ensemble technique, and a specific approach to feature selection to develop models. Using different configurations, the results are compared one against another. All of these are done after applying the necessary data prepossessing and feature engineering steps.
Machine learning has been a subject undergoing intense study across many different industries and fortunately, companies are becoming gradually more aware of the various machine learning approaches to solve their problems. However, in or- der to to fully harvest the potential of different machine learning models and to achieve efficient results, one needs to have a good understanding of the application of the models and of the nature of data. This thesis aims to investigate different approaches to obtain good results of the machine learning algorithms applied for a given prediction task. To this end the thesis proposes and implements a four different algorithms, a stacking ensemble technique, and a specific approach to feature selection to develop models. Using different configurations, the results are compared one against another. All of these are done after applying the necessary data prepossessing and feature engineering steps.
Description
Keywords
Machine Learning, Prediction, Sales, Feature Selection, Feature Engineering, Makine Öğrenimi, Tahmin, Satışlar, Özellik Seçimi, Özellik Mühendisliği