The triangular filter is a linear filter that is usually used as a smoother. The output of the rectangular filter at the moment is the weighted mean of the input values at the adjacent moments of discrete time 
.
In contrast to the rectangular filter, the weights are not constant within the time “window” 
, but decline linearly – the farther the time moment from , the smaller the weight, and the time moments that are further than the half-size 
of the filter window have zero weight.
If to plot the weights against , then the plot looks like a symmetrical triangle. This explains the “triangular” in the name of this filter.
The triangular filter is completely defined by a single parameter – the window size. The greater the parameter, the greater the degree of smoothing and, hence, the greater the suppression of higher frequencies.
As compared to the rectangular filter, the triangular filter provides better suppression of the higher frequencies and its output looks smoother.
See also the short course Time Series Forecasting.