The earliness–tardiness (E/T) measure has been the subject of abundant scheduling research efforts due to the increasing business emphasis on commitment to due dates as a competitive advantage. A notable feature of modern manufacturing is the extensive use of the Just-in-Time concept in inventory/production management where each job should be completed as close as possible to its due date. In such an environment, tardiness or earliness is important. A few applications of JIT to minimize sub-quality or defective products include the manufacturing of plastics, glass, chemicals, and semiconductors. In these manufacturing activities, raw material or work-in-process items may either degrade in quality or become unusable due to either early or late processing. This need for on-time processing i also encountered i semiconductor manufacturing, where some chemical coating processe of the wafer surface cannot be done either early or late because it may result i degrading the wafer surface. In this situation, earliness and tardiness are considered as equally undesirable. Thus, any schedule of a set of jobs should strive to minimize the total earliness and tardiness with tardiness reflecting customer satisfaction while earliness measures inventory performance. Since the problem justify; MARGIN: 0in 0in 0pt -2.85pt; unicode-bidi: embed; DIRECTION: ltr; tab-stops: -14.2pt 0in" In this thesis, we present a novel two-stage heuristic approach to minimize different early/tardy scheduling problems on a single machine. Performance is measured by the minimization of different types of earliness/tardiness costs i.e. sum of maximum earliness and tardiness, weighted versions of the total earliness and tardiness. Total tardiness is also considered as an important objective function. In the first stage of the presented approach, after generating an initial random sample, a job-position matrix is constructed. Based on this matrix and according to the procedure described in detail in related section, the approximate position of each job in a sequence is specified. The induced knowledge is applied in generating a new sample. Another job-position matrix is reconstructed. In order to find the optimal assignment of each job to each position, a Hungarian algorithm is applied to this matrix. To improve the solution quality a local search is applied. The efficiency of the proposed approach has been demonstrated by numerical results on related benchmark problems. Computational results reflect the optimal and near optimal solutions on different performance measures in a few seconds. Further work could consist in enhancing the problem to multi-operation scheduling problems such as shop scheduling problems. Applying the novel two-stage proposed method to different other objective functions such as weighted version of our problem is also worth pursuing.