Anomaly detection of cigarette sales amount and average price of total company, a specific brand and a specific salesman are important for detecting and preventing the illegal circulation of cigarette. By analyzing large amount historical sales records, a cigarette sales foresting model is built to predict the monthly and weekly sales status based on ARIMA model. And then an anomaly detective model is constructed on the predictive result. Although the ARIMA model has been analyzed extensively by researchers and used widely by forecasting practitioners due to its attractive theoretical properties and empirical evidence in its support, there are no empirical investigations have been conducted in the anomaly detection of cigarette sales. Result on the five years sales data from a Chinese city indicate that anomaly detection based on ARIMA on lunar calendar can be an effective way to improve forecasting accuracy and then improved anomaly detection accuracy is achieved.