Time series predicting is a process of analysing time series data and predicting unknown values with the help of historical data. The predictions are most often made with deep learning models and neural networks. The goal of this thesis is to build a simple neural network for predicting a single output parameter from one or more input parameters. For aiding the creation process and testing purposes, two types of data are used. The first set represents the time series of coordinates of an autonomous robot in two-dimensional space. This data is mainly used to develop a basic prediction system and for analysing the used parameters and methods. The second set represents the course of yearly weather conditions around Ljubljana. Using this data, we try to predict the daily flow of temperatures. The development of the required system as well as the predictions of all types of data were successful.
|
