Using wavelet analysis, wavelet based Independent Component Analysis (WLICA) and wavelet based Principal Component Analysis (WLPCA) (also called Emperical Orthogonal Function analysis, EOF) on individual wavelet scale power and averaged wavelet scale power (SAWP), we objectively classified East Africa into zones of coherent variability and predictability and also established links of teleconnectivity between the East Africa SON and MAM rainfall and the Indian and the south Atlantic Ocean sea surface temperatures (SST). A combined Wavelet/EOF(ICA) analysis classified East Africa into a single SON region-wide signal, while in the MAM season East Africa was classified into two out-of-phase regional signals in total contrast to the 6 or more than 20 homogenous rainfall zones as found by (Ntale, 2001) and Ogallo (1989) who used EOF on East African rainfall data. Wavelet/EOF(ICA) analysis revealed that (1) East Africa has suffered from a consistent decrease of the SON rainfall from 1960 to 1997: (2) The East Africa MAM suffered a consistent decreased of rainfall from 1982-1997 and that the failure of the MAM rains triggered the most severe droughts in East Africa, and that (3) The SON is strongly linked to the SW Indian ocean SST and less strongly to the south Atlantic ocean SST, while the MAM rainfall is strongly linked to the NW Indian Ocean SST and the Brazil and Guinea ocean currents SST in the south Atlantic ocean. Using predictors identified in the April-May-June season from the Indian and South Atlantic Ocean, excellent prediction skill depicting the strength of the identified signals is recorded over east Africa, for the SON season (2-months lead time) and MAM (8-months lead time) by a non-linear model, ANN-GA, and modest skill by a linear model, CCA.