Analisis Perbandingan Model Matrix Factorization dan K-Nearest Neighbor dalam Mesin Rekomendasi Collaborative Berbasis Prediksi Rating

Rating is a form of assessment of the likes or dislikes of a user or customer for an item. Where the higher the rating number given, the item is preferred by customers or users. In the recommendation engine, a set of ratings can be predicted and used as an object to generate a recommendation by the Collaborative Filtering method. In the Collaborative Filtering method, there is a rating prediction model, namely the Matrix Factorization and K-Nearest Neighbor models. This study analyzes the comparison of the two prediction models based on the value of Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and the prediction results generated using the movielens film rating dataset. From the analysis and testing results, it was found that MAE = 0.6371 and RMSE = 0.8305 for the Matrix Factorization model, while MAE = 0.6742 and RMSE = 0.8863 for the K-Nearest Neighbor model. The best model is Matrix Factorization because the MAE and RMSE values are lower than the K-Nearest Neighbor model and have the closest predicted rating results from the original rating value.

Source Apportionment of Particulate Matter in Urban Snowpack Using End-Member Mixing Analysis and Positive Matrix Factorization Model

The aim of this study was to identify particulate matter (PM) sources and to evaluate their contributions to PM in the snowpack of three East Siberian cities. That was the first time when the PM accumulated in the snowpack during the winter was used as the object for source apportionment study in urban environment. The use of long-term integrated PM samples allowed to exclude the influence of short-term weather conditions and anthropogenic activities on PM chemistry. To ascertain the real number of PM sources and their contributions to air pollution the results of source apportionment using positive matrix factorization model (PMF) were for the first time compared to the results obtained using end-member mixing analysis (EMMA). It was found that Si, Fe and Ca were the tracers of aluminosilicates, non-exhaust traffic emissions and concrete deterioration respectively. Aluminum was found to be the tracer of both fossil fuel combustion and aluminum production. The results obtained using EMMA were in good agreement with those obtained using PMF. However, in some cases, the non-point sources identified using PMF were the combinations of two single non-point sources identified using EMMA, whereas the non-point sources identified using EMMA were split by PMF into two single non-point sources. The point sources were clearly identified using both techniques.

Bioaerosols and dust are the dominant sources of organic P in atmospheric particles

Several studies assessed the impact of inorganic P in fertilizing oligotrophic areas, however, the importance of organic P in such fertilization processes received far less attention. In this study, the amount and origin of organic P delivered to the eastern Mediterranean Sea were characterized in atmospheric particles using the positive matrix factorization model (PMF). Phospholipids together with other chemical compounds (sugars, metals) were used as tracers in PMF. The model revealed that dominant sources of organic P are bioaerosols and dust. The amount of organic P from bioaerosols (~4 Gg P y−1) is similar to the amount of soluble inorganic P originating from dust aerosols; this is especially true during highly stratified periods when surface waters are strongly P-limited. The deposition of organic P from bioaerosols can constitute a considerable flux of bioavailable P—even during periods of dust episodes, implying that airborne biological particles can potentially fertilize marine ecosystems.

A 110 Year Sediment Record of Polycyclic Aromatic Hydrocarbons Related to Economic Development and Energy Consumption in Dongping Lake, North China

A sedimentary record of the 16 polycyclic aromatic hydrocarbon (PAH) pollutants from Dongping Lake, north China, is presented in this study. The influence of regional energy structure changes for 2–6-ring PAHs was investigated, in order to assess their sources and the impact of socioeconomic developments on the observed changes in concentration over time. The concentration of the ΣPAH16 ranged from 77.6 to 628.0 ng/g. Prior to the 1970s, the relatively low concentration of ΣPAH16 and the average presence of 44.4% 2,3-ring PAHs indicated that pyrogenic combustion from grass, wood, and coal was the main source of PAHs. The rapid increase in the concentration of 2,3-ring PAHs between the 1970s and 2006 was attributed to the growth of the urban population and the coal consumption, following the implementation of the Reform and Open Policy in 1978. The source apportionment, which was assessed using a positive matrix factorization model, revealed that coal combustion was the most important regional source of PAHs pollution (>51.0%). The PAHs were mainly transported to the site from the surrounding regions by atmospheric deposition rather than direct discharge.

Evolutionary Social Poisson Factorizationfor Temporal Recommendation

Preference-based recommendation systems analyze user-item interactions to reveal latent factors that explain our latent preferences for items and form personalized recommendations based on the behavior of others with similar tastes. Most of the works in the recommendation systems literature have been developed under the assumption that user preference is a static pattern, although user preferences and item attributes may be changed through time. To achieve this goal, we develop an Evolutionary Social Poisson Factorization (EPF$$\_$$ _ Social) model, a new Bayesian factorization model that can effectively model the smoothly drifting latent factors using Conjugate Gamma–Markov chains. Otherwise, EPF$$\_$$ _ Social can obtain the impact of friends on social network for user’ latent preferences. We studied our models with two large real-world datasets, and demonstrated that our model gives better predictive performance than state-of-the-art static factorization models.

Machine learning for Drug-Virus Prediction

The 2019 Coronavirus (COVID-19) epidemic has recently hit most countries hard. Therefore, many researchers around the world are looking for a way to control this virus. Examining existing medications and using them to prevent this epidemic can be helpful. Drug repositioning solutions can be effective because designing and discovering a drug can be very time-consuming. In this study, we used a binary classifier learning method to predict the drug-virus relationship. The feature vector for each drug-virus pair is based on the similarity between drugs and the similarity between viruses. We calculated the similarities between the drugs using their structural properties (fingerprint) and their phenotype. We also calculated the similarities between viruses based on their genome sequence and the vector encoded by the Biobert model. Finally, using the HDVD dataset, we formed the similarity vectors of each drug-virus pair and considered it as input to neural network and random forest models. In these models, we randomly selected 20% of the positive data and the same amount of negative data. Finally, the performance of the proposed approach for this test data is considered, after five tests, as AUC=0.97 and AUPR = 0.96. We also used the Compressed Sensing (CS) matrix factorization model to predict the drug-virus association. After that, we investigated the importance of drug features in predicting drug-virus association by using Autoencoder and reducing the dimension of drug properties.