Analysis of zen Price's time series
This analysis was carried out in the context of the thesis "Time series analysis for price recommendation in the telecommunications market", whose main objective was to develop statistical techniques and methods to make price predictions of the prices of products in the telecommunications market. The time series used in this work were obtained by zen Price, which is a Saas solution that collects the prices of e-commerce products via web-scrapping.
This repository contains several Jupyter notebooks containing the analysis of these time series. The Jupyter notebooks are divided by the chapters of the dissertation. A brief description of the purpose of each chapter follows.
Chapter 2 - introduction to the data and brief exploratory analysis of the time series, which allowed characterizing the time series (length of the time series and number of price changes per time series), identifying time patterns, dealing with missing values...
Chapter 3 - use classical techniques to forecast prices. A set of naive models and ARIMA-like models were tested.
Chapter 4 - use the Prophet tool, create by Facebook, to forecast prices.
Chapter 5 - forecast price changes (using Markov Chains, Logistic Regression, LSTM, Random Forest and SVM models) and study techniques (e.g. Dynamic Time Warping) for identifying time-varying lead-lad relationships in groups of time series.
The master's thesis can be consulted at xxxxxx.
For more information consult:
Vitória Cruz, Author of the dissertation, vicruz99@ua.pt
Sónia Gouveia, PhD Researcher, Supervisor of the dissertation, sonia.gouveia@ua.pt.