Kolmogorov-Arnold Networks (KAN) are a novel neural network architecture inspired by the Kolmogorov-Arnold representation theorem. This paper explores the application of KAN to time series forecasting, introducing two variants: Temporal KAN (T-KAN) and Multivariate Temporal KAN (MT-KAN). T-KAN is designed to detect and track concept drift in univariate time series, enhancing interpretability through symbolic regression. MT-KAN improves predictive performance by leveraging complex relationships among variables in multivariate time series. Experiments on financial time series data show that both T-KAN and MT-KAN achieve comparable or superior results with fewer parameters compared to traditional models like MLPs, RNNs, and LSTMs. The paper highlights the potential of KAN as a powerful and interpretable tool in predictive analytics, particularly in dynamic and complex environments. Future work will focus on integrating KAN with other architectures and optimizing training speed for real-time applications.Kolmogorov-Arnold Networks (KAN) are a novel neural network architecture inspired by the Kolmogorov-Arnold representation theorem. This paper explores the application of KAN to time series forecasting, introducing two variants: Temporal KAN (T-KAN) and Multivariate Temporal KAN (MT-KAN). T-KAN is designed to detect and track concept drift in univariate time series, enhancing interpretability through symbolic regression. MT-KAN improves predictive performance by leveraging complex relationships among variables in multivariate time series. Experiments on financial time series data show that both T-KAN and MT-KAN achieve comparable or superior results with fewer parameters compared to traditional models like MLPs, RNNs, and LSTMs. The paper highlights the potential of KAN as a powerful and interpretable tool in predictive analytics, particularly in dynamic and complex environments. Future work will focus on integrating KAN with other architectures and optimizing training speed for real-time applications.