Demand forecasting supports the business to arrive at more intelligent supply decisions by estimating future sales and revenue. By anticipating future sales, it enables organizations to optimize inventories. Demand managers may make well-informed company decisions about all that from planning for supplies and warehousing requirements to running quick sales and satisfying consumer expectations by examining past sales data. The primary objective is to build a model using light gradient boosting machine regression that precisely predicts sales for the upcoming year using machine learning. This model renders great interpretability and high efficiency, making it ideal for use in industrial production environments. In order to estimate sales for the upcoming 12 months for 50 distinct items across ten different stores, sales data for retail items from the last five years have been gathered. The model utilizes past transaction data to forecast upcoming daily sales. The data comes with an upward trend and certainly seems seasonal. The assumption is further supported by the fact that the mean or average of the residuals appears to be zero. A continuous forecast is updated periodically throughout the year, which predicts a company's performance over an ongoing period of time. Date-time features are properties that are derived from the time stamp value of each observation. The data have been split into both the train and the test sets, in which the train set including sales data from the past five years and the test set including sales from the last twelve months. For the performance measurement of errors, the following metrics are evaluated: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and R2-squared using grid search model and have validated our cross-validation score. The information from this analysis has given the vendor the ability to foresee and estimate future demand from customers for a specific service or product.