The famer-herder crisis over scarce grazing land resources has over the years resulted to the loss of several lives and properties as well as reduction in food production in Benue State. Social media plays a key role in conflict escalation or de-escalation because they are information carriers. In this study, Sentiment analysis is used to analyse opinions of social media comments and posts about farmers-herders crisis expressed by Twitter users to ascertain the positive, negative and neutral tweets. Tweets were extracted using Python library called Snscrape, model was designed and processing was done using the Natural Language Tool Kit (NLTK) to categorize the data into positive, negative and neutral. The naïve Bayes and random forest were used for the analysis and evaluation was done using Receiver Operating Characteristics (ROC), F1 and Accuracy. The values for ROC, F1 and Accuracy for Random Forest and Naïve Bayes algorithms are 0.73, 0.61 and 0.90, and 0.70, 0.73 and 0.60 respectively. It was observed that Random Forest performed better than Naïve Bayes, and as such, can be applied towards assisting security agencies and government policies on the crisis management.