<p>Uncertainties in the system are likely not to be handled simultaneously due to randomness and fuzziness whereas probabilistic fuzzy set has been proved by the researchers to model such kind of uncertainties in different domains. In this paper, we propose a cumulative probability based discretization and particle swarm optimization based probabilistic fuzzy time series forecasting method. Gaussian probability distribution function has been used to evaluate the probabilities and associates to membership grades then constructed probabilistic fuzzy set. Advantage of the proposed work is that it may handle the uncertainties due to randomness and fuzziness in a single framework, it also improves accuracy rate in time series forecasting. Particle swarm optimization used to optimize the length of intervals which are obtained by cumulative probability based discretization approach and probabilistic fuzzy set can handle both kinds of uncertainties stochastic and non-stochastic simultaneously. In proposed forecasting method, we are also using an aggregation operator to aggregate probabilistic fuzzy outputs to fuzzy row vector. Proposed forecasting method is applied on time series data set of enrolments of University of Alabama and two financial datasets of State Bank of India share prices at Bombay Stock Exchange, India and Taiwan Stock Exchange Capitalization Weighted Stock Index. Out performance of proposed forecasting method based on probabilistic fuzzy set and particle swarm optimization is verify by using root mean square error, mean absolute deviation, tracking signal and evaluation parameter.</p>