Random Forest Analysis of Cassava Mosaic Disease Factors: Impacts on Livelihoods of Cassava Farmers

Cassava mosaic disease (CMD) is a significant threat to cassava production in various regions, particularly in Africa. The disease is caused by several geminiviruses, including African cassava mosaic virus (ACMV), East African cassava mosaic virus (EACMV), and East African cassava mosaic Cameroon virus (EACMCMV). This study aimed to identify key CMD-related factors affecting Cameroon cassava farmers’ incomes originating from both the sale of cassava cuttings (V215) and the sale of cassava roots (V216). The dataset had 630 records originating from respondents residing within 4 regions of Cameroon – Adamawa, Center, East and South. Here, nine CMD-related variables were used to independently train two Random Forest models. Random forest residual plots are a crucial tool in model exploration and diagnostics. These models were later employed for regression-based prediction of both financial targets V215 and V216. The Random Forest (RF)-based mean absolute percentage error for targets V215 and V216 were 0.19 and 1.25 respectively. The RF-based mean Gaussian deviance for targets V215 and V216 were 0.07 and 0.51 respectively. Based on RF feature importance scores (RFFI), the top 3 factors affecting income originating from the sale of cassava cuttings were found to be: late appearance of symptoms as a difficulty associated with regular field monitoring (RFFI of 0.2594), removal of infected plants as a method of controlling frequent occurrence of viral diseases in respondents’ cassava fields (RFFI of 0.1633) and lack of healthy planting material due to frequent occurrence of viral diseases in respondents’ cassava fields (RFFI of 0.1495). Also, the top 3 factors affecting income originating from the sale of cassava roots were found to be: the replacement of infected plants with healthy cuttings as a method of controlling the frequent occurrence of viral diseases in respondents’ cassava fields (RFFI of 0.1974), decrease in yield due to frequent occurrence of viral diseases in respondents’ cassava fields (RFFI of 0.1530) and poor plant growth due to frequent occurrence of viral diseases in respondents’ cassava fields (RFFI of 0.1388). This study has shown that the continuous reliance of cassava farmers on vegetative propagation-based cassava planting material is principally responsible for the bulk of the financial losses they experience, and so an adoption of seed-based cassava planting material would go a long way to secure their on-farm livelihood in cassava production.