Imputation by feature importance (IBFI): A methodology to envelop machine learning method for imputing missing patterns in time series data

Binned Data Provide Better Imputation of Missing Time Series Data from Wearables

The presence of missing values in a time-series dataset is a very common and well-known problem. Various statistical and machine learning methods have been developed to overcome this problem, with the aim of filling in the missing values in the data. However, the performances of these methods vary widely, showing a high dependence on the type of data and correlations within the data. In our study, we performed some of the well-known imputation methods, such as expectation maximization, k-nearest neighbor, iterative imputer, random forest, and simple imputer, to impute missing data obtained from smart, wearable health trackers. In this manuscript, we proposed the use of data binning for imputation. We showed that the use of data binned around the missing time interval provides a better imputation than the use of a whole dataset. Imputation was performed for 15 min and 1 h of continuous missing data. We used a dataset with different bin sizes, such as 15 min, 30 min, 45 min, and 1 h, and we carried out evaluations using root mean square error (RMSE) values. We observed that the expectation maximization algorithm worked best for the use of binned data. This was followed by the simple imputer, iterative imputer, and k-nearest neighbor, whereas the random forest method had no effect on data binning during imputation. Moreover, the smallest bin sizes of 15 min and 1 h were observed to provide the lowest RMSE values for the majority of the time frames during the imputation of 15 min and 1 h of missing data, respectively. Although applicable to digital health data, we think that this method will also find applicability in other domains.

The applications of Machine learning (ML) in designing dry powder for inhalation by using thin-film-freezing technology

Dry powder inhalers (DPIs) are one of the most widely used devices for treating respiratory diseases. Thin--film--freezing (TFF) is a particle engineering technology that has been demonstrated to prepare dry powder for inhalation with enhanced physicochemical properties. Aerosol performance, which is indicated by fine particle fraction (FPF) and mass median aerodynamic diameter (MMAD), is an important consideration during the product development process. However, the conventional approach for formulation development requires many trial-and-error experiments, which is both laborious and time consuming. As a state-of-the art technique, machine learning has gained more attention in pharmaceutical science and has been widely applied in different settings. In this study, we have successfully built a prediction model for aerosol performance by using both tabular data and scanning electron microscopy (SEM) images. TFF technology was used to prepare 134 dry powder formulations which were collected as a tabular dataset. After testing many machine learning models, we determined that the Random Forest (RF) model was best for FPF prediction with a mean absolute error of ± 7.251%, and artificial neural networks (ANNs) performed the best in estimating MMAD with a mean absolute error of ± 0.393 μm. In addition, a convolutional neural network was employed for SEM image classification and has demonstrated high accuracy (>83.86%) and adaptability in predicting 316 SEM images of three different drug formulations. In conclusion, the machine learning models using both tabular data and image classification were successfully established to evaluate the aerosol performance of dry powder for inhalation. These machine learning models facilitate the product development process of dry powder for inhalation manufactured by TFF technology and have the potential to significantly reduce the product development workload. The machine learning methodology can also be applied to other formulation design and development processes in the future.