We present a statistical analysis of the long-term mid-infrared variability for WISE-selected AGN over 270 square deg. of Sloan Stripe 82 field. AGN sample is selected using 3.4 and 4.6-micron channels of WISE and contains both obscured (Type-II) and unobscured (Type-I) AGN. We use WISE W1, and W2 single exposure (Level 1b) images to construct light curves for each of these AGN with typical baselines of ~8 years. We then search for statistically significant flux variations in both the channels of WISE to identify 767 AGN as MIR variable, across a wide range of redshift 0 < z < 2.2. Overall, we estimate that ~20% of AGN are MIR variable at W1 < 17.7 AB mag. Our variability estimation is sensitivity-limited and biased towards luminous AGN. We also characterize the AGN sample based on optical photometry from the Pan-STARRS survey and spectroscopic redshifts from SDSS surveys. We further classify the AGN population as dust-obscured vs unobscured and analyze them separately. We do not find any statistically significant differences in the variability pattern of obscured and unobscured AGN populations. Our models indicate that the long-term MIR variability seen in the WISE light curves of AGN is predominantly due to variations in the accretion rate, with the energy being reprocessed by ~1000K dust in the vicinity of the AGN. We further estimate that typically ~4% of emission from the accretion disk is reprocessed by the surrounding dust in unobscured AGN.