Sales Forecast by using Convolutional Neural Networks

All companies need an effective method to predict future sales and several classic, purely statistical methods exist and are heavily utilized in the industry. However, these prediction methods are not always effective when faced with predictions with no clear trend in past data. Recently more sophisticated techniques with improved prediction capacities, based on Machine Learning and Neural Networks, have appeared. For instance, Feedforward Neural Networks and Recurrent Neural Networks have demonstrated good precision when working on time series prediction. But, both methods fail when many hidden layers are utilized, this is due to the well-known gradient vanishing problem. This work proposes a novel sales prediction method based on Convolutional Neural Networks. This type of neural network is generally used for image processing tasks. But in this work, we explore new applications and develop models that produce good results in sales prediction for real pharmaceutical product data. The used data belongs to an Ecuadorian pharmacy franchise database. It contains weekly sales of products over a period of 4 years. With this data set, several classical prediction methods and artificial intelligence prediction methods were implemented and tested. Then, our proposed method based on convolutional neural networks was designed and programmed in Matlab. After this, all these prediction methods were compared using three metrics: prediction accuracy, number of weights and number of iterations. Finally, we proceeded to determine which prediction method is better both in accuracy and efficiency as well as their advantages and disadvantages.

• Yazar: Velasteguí Sandoval, Ronny Xavier (author)
• Materyal Türü: bachelorThesis
• Dil: eng
• Baskı/Yayın Bilgisi: 2020
• Konular: Sales Forecast; Artificial Neural Networks; Convolutional Neural Networks; Deep Learning; Predicción de ventas; Redes neuronales artificiales; Redes neuronales convolucionales; Aprendizaje profundo
• Online Erişim: http://repositorio.yachaytech.edu.ec/handle/123456789/118