As temporal data are of growing importance hence require to support new trends for filtering and smoothing. The main goal of this work is to explore new smoothing techniques like wavelet filters in place of traditional moving averages and regression methods, etc for temporal data for further decision making processes like forecasting, similarities search and clustering. Temporal applications are increasingly popular methods for studying a wide range of databases including sensor, weather, stock exchange, seismic and atomic, etc. Time series, a common form of temporal data, are mostly contaminated with noise and outliers. The direct operation and analysis on these types of raw data, may easily lead to wrong decisions. The selection of appropriate smoothing techniques is the pivotal point to reach to the correct conclusions. In this work, the main task has been to select the best possible smoothing technique for selection of features of time series to perform various tasks.