This research project explores linear and non-linear links between features for predicting time series data consisting of a real estate index. The features consists of macro-economic indicators, confidence indexes, exchange rate, oil price, and historical data of the collection of stocks that make up the index.Each feature is tested for stationary using Augmented Dickey-Fuller test and differencing until the data is stationary with a level of significance of 0.05. The data is then standardised and then fed into LDA and ANN separately. The main findings are that the discriminant scores of the LDA model captured a good separation both using separation lines and an approximation of the 3-dimensional multivariate probability density function. However, the model was comparable to or worse than random chance when applied to unseen data. For the ANN method, the model improved at a concerning low and volatile, and consequentially the accuracy was remarkably lower for the test data. The greatest limitations to the project is data preprocessing and complexities for the models. Overall, this work contributes to the linkages between ANN and LDA and time series forecasting real estate stocks.