In this study, the methods used for the position/trade action (buy/sell/hold) estimates of financial assets, especially in the literature, were examined. As a result of the examinations, it was determined that the data imbalance problem arose when performing position labeling on stock price data. In this context, the positions of the four stocks in the BIST30 index after one month were estimated using the k-nearest neighbor and support vector machines methods. The data imbalance problem that occurred during the labeling of stock data as buy/sell/hold was resolved using the SMOTE approach. In addition, the effect of using various attributes based on time series characteristics in addition to price data and technical indicators to predict the position of stock data on model performance was also investigated. We created four different input sets for the prediction models in this context. In two of these sets, monthly data of stocks and 15 technical indicator values were used in addition to these data. In the other two, respectively, the daily data of stocks and, in addition to these data, statistical attributes obtained from 15 technical indicator values are discussed. The results showed that using these statistical features increased the model's performance by 15-20% and reached an Fl-score value of 0.97.