Binary Particle Swarm Optimization (BPSO) based optimal network re-configuration and Particle Swarm Optimization (PSO) based optimal distributed generation (DG) placement and sizing, is proposed for radial distribution networks with uncertainty in generation and load, using affine arithmetic (AA), to minimize real power losses, DG investment cost and improve voltage profile. Improved AA based division operation which gives one less noise term as compared to existing AA division is used in the proposed AA based methods. An affine expression for real power loss is proposed for distribution systems, to be used in the proposed AA based methods. The major contribution of the present work is the application of improved AA based optimal network reconfiguration and optimal DG allocation, in radial distribution networks with uncertainty. The proposed AA based methods are tested on IEEE 33, 85 and 119 bus distribution systems. The optimal loss intervals and voltage profile obtained by the proposed AA based optimal network reconfiguration are compared with that obtained by interval arithmetic (IA) and Monte Carlo (MC) simulations-based methods. The simulation results show that proposed AA based analysis gives a more optimal network reconfiguration, for radial distribution systems with uncertainty as compared to existing IA based method. Similarly, optimal results obtained by proposed AA based optimal DG placement and sizing algorithm is compared with existing AA based method. It is seen that the proposed AA based algorithm gives more optimal DG sizes and locations as compared to existing AA based method.