Forecasting through Classification of time series data is a major research areas in Computer Science, Economics and many other areas. A lot of literature is available on forecasting time series data through time series analysis using statistical and mathematical methods like ARIMA, spectral analysis, fuzzy sets. Another alternative for time series forecasting through data mining technique is time series classification, which has less contribution to the literature available. There are also separate methods for predicting class labels of categorical time series data and predicting values of continuous time series data. In this work, we proposed a new method called forecasting labelled and unlabeled time series data through histogram classification using Adaptive Cost Dynamic Time Warping and Euclidean Distance. Which is a common method for predicting class labels of labelled time series data and predicting values of an unlabeled time series data. We applied this method to forecast Taiwan stock indexing, which is a continuous time series data and applied to labelled time series data called BeetleFly dataset and BirdChicken dataset to predict its class labels. The accuracy of forecasting and prediction of class labels compared with existing methods. The results showed that this method has a better performance compared to existing ones.\r\n\r\nKeywords: Forecasting, Histogram-classification, Labelleddata, Unlaelled data, K nearest neighbors.\r\n

Username
Password