A Probability-Based Models Ranking Approach: An Alternative Method of Machine-Learning Model Performance Assessment.

Performance measures are crucial in selecting the best machine learning model for a given problem. Estimating classical model performance measures by subsampling methods like bagging or cross-validation has several weaknesses. The most important ones are the inability to test the significance of the difference, and the lack of interpretability. Recently proposed Elo-based Predictive Power (EPP)-a meta-measure of machine learning model performance, is an attempt to address these weaknesses. However, the EPP is based on wrong assumptions, so its estimates may not be correct. This paper introduces the Probability-based Ranking Model Approach (PMRA), which is a modified EPP approach with a correction that makes its estimates more reliable. PMRA is based on the calculation of the probability that one model achieves a better result than another one, using the Mixed Effects Logistic Regression model. The empirical analysis was carried out on a real mortgage credits dataset. The analysis included a comparison of how the PMRA and state-of-the-art k-fold cross-validation ranked the 49 machine learning models, an example application of a novel method in hyperparameters tuning problem, and a comparison of PMRA and EPP indications. PMRA gives the opportunity to compare a newly developed algorithm to state-of-the-art algorithms based on statistical criteria. It is the solution to select the best hyperparameters configuration and to formulate criteria for the continuation of the hyperparameters space search.

Prevalence of hypogammaglobulinemia and its management with subcutaneous immunoglobulin supplementation in patients after allogeneic hematopoietic stem cell transplantation-a single-center analysis.

Secondary immunodeficiencies are frequently observed after allo-HSCT. The efficacy of subcutaneous IgG preparations in this population is unknown. A retrospective single-institution study involved 126 adult patients transplanted in 2012-2019 for hematological malignancies. Patients were tested every 2-3 weeks for plasma IgG concentration during the 1st year after transplantation and supplemented with facilitated subcutaneous immunoglobulin when they either had IgG concentration < 500 mg/dl or between 500 and 700 mg/dl and recurrent infection. The IgG concentration < 500 mg/dL was diagnosed in 41 patients, while 500-700 mg/dL in 25 and altogether 53 patients received IgG supplementation. The median number of IgG administrations was 2. The median time to the first IgG administration after allo-HSCT was 4.1 months, while to the next administration (if more than one was required) 53 days (prophylactic group) and 32 days (group with infections). We did not observe any significant toxicity. Two situations were associated with increased probability of meeting criteria for IgG supplementation: diagnosis of either acute lymphoblastic leukemia (ALL) or chronic lymphocytic leukemia (CLL) (83.8% versus 39.3% for other diagnosis, p = 0.000) and the systemic use of corticosteroids (64.2% versus 31.5% for patients without systemic corticosteroids, p = 0.005). Over 40% of the adult recipients may require at least incidental immunoglobulin supplementation during the first year after allo-HSCT. Low IgG concentrations are associated with inferior outcomes. The subcutaneous route of IgG administration appeared to be safe and may allow for long persistence.