Determining the values of movable assets at specific terms using hybrid artificial intelligence methods

The increasing volatility of financial markets, in addition to the limitations of pure traditional methods of recognition, justifies that there is a demand to employ systems that are more advanced and data-driven experimental approaches for assessing mutual funds' performance. On this basis, the research proposes that mutual funds in Turkey could be studied by the means of technical analysis indicators together with AI-backed predictive models. The study referred to two major datasets (Equity Umbrella Funds vs. Precious Metal Umbrella Funds). Data was obtained from credible sources, i.e., Takasbank and Public Disclosure Platform (KAP), including components such as past fund prices, daily returns, investor count, and total value of portfolios. The study was conducted in two phases. In the first step, the market participation of the fund was represented by technical indicators (moving averages - MA20 and MA50 -, the volatility by STDDEV20, or the correlations matrix). The second phase of the research consisted of the construction of a model for short-term predictability of the price of mutual funds using LSTM (Long Short-term Memory), a critical approach to AI forecast. Data preprocessing included Min-Max normalization, and the model's performance was measured using metrics such as MSE, MAE, and R-squared. The 5-fold cross-validation was used to enhance the generalization of the model. Results suggested that some of the funds were forecastable by using TA or LSTM models, we observed similar performance, while the rest were more volatile than others, which means more risk for the funds. Consequently, this thesis shows investors, portfolio managers, and researchers how to combine two concepts of financial analysis, technical and predictive analysis and offers a new point of view for the previous literature.