The Intelligent System for Interactive Analysis and Forecasting of Graph Data

Moshkin V., Yarushkina N., Moshkina I.

4th International Conference on Intelligent and Fuzzy Systems (INFUS), Bornova, Turkey, 19 - 21 July 2022, vol.504, pp.870-878 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 504
  • Doi Number: 10.1007/978-3-031-09173-5_100
  • City: Bornova
  • Country: Turkey
  • Page Numbers: pp.870-878
  • Keywords: Machine vision, Forecasting, Time series, Neural network
  • Yıldız Technical University Affiliated: No


The article describes a mobile software system that recognizes function graphs using computer vision and machine learning methods, their analysis and prediction using several intelligent algorithms. The software system recognizes the graph in the photograph, determines the numerical values of the graph points, predicts the resulting time series and draws the continuation of the graph. Hough Transform is used to recognize graphs, and several models were used to predict time series: linear regression, ARIMA, S-model and a neural network of our own architecture using keras. Three function graphs were recognized to select an efficient forecasting approach. The RMSE was used to assess the effectiveness of forecasting. Experiments have shown that for different types of charts, different methods are needed; different forecasting methods are effective. In this regard, it is necessary to implement a method for selecting a forecasting model for a specific type of recognizable graph.