Time Series Forecasting : Developing an AI-Integrated Web Server for Forecasting
Chen, Zenan (2024)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2024122538128
https://urn.fi/URN:NBN:fi:amk-2024122538128
Tiivistelmä
In today’s fast-paced, data-driven world, businesses and organizations often face inefficiencies and wasted resources. Issues such as overstocking, underutilized energy systems, and inaccurate demand forecasts for public places can lead to increased costs and operational setbacks. Is it possible to develop a web server with integrated AI models that allows businesses and organizations to plan more effectively, predict trends more accurately, and reduce inefficiencies and wasted resources in a convenient and fast way?
This thesis explores the integration of AI models in a web server to implement a web server that can automatically predict time series data. The web server will have two main functions, one is to automatically process and train data, and the other is to predict data and display the predicted data in a dynamic form.
Finally, a web server that can automatically predict and visualize time series data was implemented. This thesis details the construction of the web server, the integration of models, the challenges encountered, and suggestions for improvement. This web server provides a practical and scalable solution for industries requiring time-sensitive insights, such as energy management, retail, and logistics.
Through an in-depth study of traditional machine learning models and deep learning models, it is found that traditional machine learning is not only more accurate (93% accuracy), but also has a shorter training time, making it more suitable as an AI model integration solution for this server. However, due to its limitations, deep learning models are not ideal in terms of accuracy or training time, so they are not suitable for integration in this server.
This thesis explores the integration of AI models in a web server to implement a web server that can automatically predict time series data. The web server will have two main functions, one is to automatically process and train data, and the other is to predict data and display the predicted data in a dynamic form.
Finally, a web server that can automatically predict and visualize time series data was implemented. This thesis details the construction of the web server, the integration of models, the challenges encountered, and suggestions for improvement. This web server provides a practical and scalable solution for industries requiring time-sensitive insights, such as energy management, retail, and logistics.
Through an in-depth study of traditional machine learning models and deep learning models, it is found that traditional machine learning is not only more accurate (93% accuracy), but also has a shorter training time, making it more suitable as an AI model integration solution for this server. However, due to its limitations, deep learning models are not ideal in terms of accuracy or training time, so they are not suitable for integration in this server.