In this research analysis, in addition to the conventional ARIMA model, specific long short-term memory (LSTM), stacked-LSTM, and concept-based LSTM were used to calculate next day expenses. Furthermore, using our expectation, we developed two modes of transmission, different and important identity. Our database information not only includes the usual end-of-day expense and transfer modules, but also includes corporate bookkeeping insights, which are effortlessly selected and used in samples. With the regular ARIMA model, learning the next model in anticipation of the next day's stock costs, especially long short-term memory model (LSTM), stacked-LSTM, and concept-based LSTM. In addition, using our forecast, we developed two exchange procedures and developed differential and scale. Our database information not only includes regular end-of-day cost and transfer modules, but also includes corporate bookkeeping metrics, which are deliberately selected and used in samples. Bookkeeping information is considered information and cost plans for a company that no longer relies on expanding the pioneering power of the model. The effect indicates that the LSTM beats any remaining model in relation to the forecast error and shows a lot better yield in our transfer practice on different models. Besides, we found that the stacked-LSTM model does not improve the advance control over the LSTM.