A sequential decision-making model was developed, and data from farm-raised catfish production were used to demonstrate its use. Outcomes of sequences of decisions which satisfied chance constraints on ending cash balances were traced through a specified time period. Discrete choice variables were specified due to the fixed nature of pond facilities. Recourse actions specified were sale of production in excess of endogenously determined transfer levels or purchase of inputs to supplement needs of the next production stage. Production activities cannot be changed during the planning period. Only yield variability was considered due to its impact on relative competitiveness among growth stages. Deviations were calculated from endogenously determined target levels based on goal and probability limits.

This paper aims at proposing tractable algorithms to find effectively good solutions to large size chance-constrained combinatorial problems. A new robust model is introduced to deal with uncertainty in mixed-integer linear problems. It is shown to be strongly related to chance-constrained programming when considering pure 0–1 problems. Furthermore, its tractability is highlighted. Then, an optimization algorithm is designed to provide possibly good solutions to chance-constrained combinatorial problems. This approach is numerically tested on knapsack and multi-dimensional knapsack problems. The results obtained outperform many methods based on earlier literature.