RESUMO
The linear regression is critical for data modelling, especially for scientists. Nevertheless, with the plenty of high-dimensional data, there are data with more explanatory variables than the number of observations. In such circumstances, traditional approaches fail. This paper proposes a modified sparse regression model that solves the problem of heterogeneity using seaweed big data as a use case. The modified heterogeneity models for ridge, LASSO and Elastic net were used to model the data. Robust estimations M Bi-Square, M Hampel, M Huber, MM and S were used. Based on the results, the hybrid model of sparse regression for before, after, and modified heterogeneity robust regression with the 45 high ranking variables and a 2-sigma limit can be used efficiently and effectively to reduce the outliers. The obtained results confirm that the hybrid model of the modified sparse LASSO with the M Bi-Square estimator for the 45 high ranking parameters performed better compared with other existing methods.
RESUMO
Political speech acts are critical for politicians launching a regime because they can provide information that can be used to control people's thoughts and opinions. The purpose of this study was to conduct a qualitative content analysis of the inaugural and ascension addresses of Nigerian heads of state and presidents. The textual data used in this analysis were the ascension and inaugural addresses of Nigerian Heads of State and Presidents from 1960 to 2019. They were extracted and analysed using text-mining techniques. Textual data were clustered about their topical content using Latent Dirichlet Allocation (LDA), and speech cohesion between these addresses was examined using a similarity matrix and heatmap. Furthermore, term frequency and association analyses were performed to examine the high-frequency terms (tokens) and the terms (tokens) that are strongly correlated within each of the ascension/inaugural addresses (corpus). The summarization of characters and words in the ascension and inaugural addresses reveals that the Civilian Presidents used more characters and words than the Military Heads of State. There was an increase in the number of characters and words in the ascension and inaugural addresses among those who had served the nation multiple times. The total sentiment score in the ascension/inaugural addresses from 1960 to 2019 by Civilian Presidents and Military Heads of State revealed that the Civilian Presidents expressed more trust, surprise, sadness, joy, fear, disgust and anticipation in their addresses than the Military Heads of State. The most occurring term (token) in the ascension/inaugural addresses was the word government which appeared 221 times. The most token in the corpus government was found to be moderately correlated with the following tokens: loss, existing and majority. Similarly, economic was found to be moderately correlated with these tokens: inflation, building, education, exchange, loan, workers and technical. In this study, all the ascension/inaugural addresses share similar topic distribution: as seen in Abacha's and Muritala's addresses; and Shonekan's inaugural address was very similar to Balewa, Azikwe and Babangida's addresses; Babangida's ascension, Abdulsalam's 1998 ascension, Jonathan's 2010 inaugural and Buhari's 2015 inaugural addresses discussed similar topics to Obasanjo's 1976 ascension address. The highest average sentiment score was observed in Obasanjo's 2003 inaugural address and the lowest score was in Buhari's 1983 ascension address. The sentiment score for the ascension/inaugural addresses showed that Civilian Presidents inaugural addresses expressed more positive, joy, trust and anticipation than Military Heads of State. These emotions showed that the Civilian President's inaugural addresses are better when compared to Military Heads of State in terms of the sentiment scores.