The effectiveness of information retrieval systems heavily depends on a large number of hyperparameters that need to be tuned. Hyperparameters range from the choice of different system components, e.g., stopword lists, stemming methods, or retrieval models, to model parameters, such as k1 and b in BM25, or the number of query expansion terms. Grid and random search, the dominant methods to search for the optimal system configuration, lack a search strategy that can guide them in the hyperparameter space. This makes them inefficient and ineffective. In this paper, we propose to use Bayesian Optimization to jointly search and optimize over the hyperparameter space. Bayesian Optimization, a sequential decision making method, suggests the next most promising configuration to be tested on the basis of the retrieval effectiveness of configurations that have been examined so far. To demonstrate the efficiency and effectiveness of Bayesian Optimization we conduct experiments on TREC collections, and show that Bayesian Optimization outperforms manual tuning, grid search and random search, both in terms of retrieval effectiveness of the configuration found, and in terms of efficiency in finding this configuration. CCS CONCEPTS• Information systems → Retrieval models and ranking;

To improve the accuracy and reliability of short-term power load forecasting and reduce the difficulty caused by load volatility and non-linearity, a hybrid forecasting model (CEEMDAN-SE-VMD-PSR-WOA-SVR) is proposed. Firstly, the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is employed to generate multiple intrinsic modal functions (IMF) by decomposing the historical power load series. Then the sample entropy (SE) of each IMF is calculated to quantitatively evaluate the corresponding complexity. Afterward, variational mode decomposition (VMD) is adopted to achieve secondary decomposition for the component with the maximum sample entropy. Subsequently, the phase space reconstruction (PSR) is applied to reconstruct each IMF into a high-dimensional feature space matrix, which is formed as the input of support vector regression (SVR). Finally, SVR optimized by whale optimization algorithm (WOA) is used for the prediction, where the predicted values of all IMFs are accumulated to obtain the final prediction results. The experimental result demonstrates that the proposed hybrid model can effectively decompose the load series with non-linear characteristic and provide more accurate forecasting results by comparing the other models. INDEX TERMS short-term power load forecasting, two-phase decomposition, sample entropy, whale optimization algorithm, support vector regression

No abstract