Nowadays usage of computers and hand held devices in our day to day life for data processing is a common task. For processing texts need to enter to the computer via keyboard or other mechanisms. Predicting a word at the beginning is one of the techniques that facilitates the data entry to the computer. Proposing a word that the user intends to type based on the context or word dictionary is the main task of word autocomplete and prediction, which is the main aim of our research. And the prediction can be used further researches as well as to support other applications like handwriting recognition, mobile phone or PDA texting, and assisting people with disabilities. A detailed study on the different factors that affect the quantitative evaluation of the data (where a normalization effort should be done) has been done. To propose a solution, we used back off mechanism to predict a word, based on the given input with the n-gram based prediction. The goal of this study was to design a new word predictor for Tigrigna text entry that would suggest words that are more grammatically appropriate based on the given input. The study has saved significantly more keystrokes. The new predictor model that we designed used a probabilistic language model based on the well-established data of the combination of unigram, bigram, trigram and quad gram predictor. In this study, the test has been shown a result in keystroke savings of 55.06%. So the study minimize the time consumed to write a document in Tigrigna by more than 50%. And that helps peoples who are not rich in vocabulary by avoiding spelling errors. Yet there are some challenges to predict word in Tigrigna. In the data analysis part of the study some affixes are difficult to differentiate. Because sometimes they are a words by themselves and sometime they are affixes.