In a public procurement process, corruption can occur at each stage, favoring a participant with a previous agreement, which can result in over-pricing and purchases of substandard products, as well as gender discrimination. This paper’s aim is to detect biased purchases using a Spanish Language corpus, analyzing text from the questions and answers registry platform by applicants in a public procurement process in Ecuador. Additionally, gender bias is detected, promoting both men and women to participate under the same conditions. In order to detect gender bias and favoritism towards certain providers by contracting entities, the study proposes a unique hybrid model that combines Artificial Intelligence algorithms and Natural Language Processing (NLP). In the experimental work, 303,076 public procurement processes have been analyzed over 10 years (since 2010) with 1,009,739 questions and answers to suppliers and public institutions in each process. Gender bias and favoritism were analyzed using a Word2vec model with word embedding, as well as sentiment analysis of the questions and answers using the VADER algorithm. In 32% of cases (96,984 answers), there was favoritism or gender bias as evidenced by responses from contracting entities. The proposed model provides accuracy rates of 88% for detecting favoritism, and 90% for detecting gender bias. Consequently one-third of the procurement processes carried out by the state have indications of corruption and bias. In Latin America, government corruption is one of the most significant challenges, making the resulting classifier useful for detecting bias and favoritism in public procurement processes.

: Shortages of essential supplies used to prevent, diagnose, and treat COVID-19 have been a global concern, and price speculation and hikes may have negatively influenced access. This study identifies variability in prices of products acquired through government-driven contracts in Ecuador during the early pandemic response, when the highest mortality rates were registered in a single day. Data were obtained from the National Public Procurement Service (SERCOP) database between March 1 and July 31, 2020. A statistical descriptive analysis was conducted to extract relevant measures for commonly purchased products, medical devices, pharmaceutical drugs, and other goods. Among the most frequently purchased products, the greatest amounts were spent on face masks (US$4.5 million), acetaminophen (US$2.2 million), and reverse transcriptase quantitative polymerase chain reaction assay kits (US$1.8 million). Prices varied greatly, depending on each individual contract and on the number of units purchased; some were exceptionally higher than their market value. Compared with 2019, the mean price of medical examination gloves increased up to 1,307%, acetaminophen 500 mg pills, up to 796%, and oxygen flasks, 30.8%. In a context of budgetary constraints that actually required an effective use of available funds, speculative price hikes may have limited patient access to health care and the protection of the general population and health care workers. COVID-19 vaccine allocations to privileged individuals have also been widely reported. Price caps and other forms of regulation, as well as greater scrutiny and transparency of government-driven purchases, and investment in local production, are warranted in Ecuador for improved infectious disease prevention.

The awarding of public procurement processes is one of the main causes of corruption in governments, due to the fact that in many cases, contracts are awarded to previously agreed suppliers (favouritism); for this selection, the qualification parameters of a process play a fundamental role, seeing as due to their manipulation, bidders with high prices win, causing prejudice to the state. This study identifies processes with anomalies and generates a model for detecting possible corruption in the assignment of process qualification parameters in public procurement. A multi-phase model was used (the identification of anomalies and generation of the detection model), which uses different algorithms, such as clustering (K-Means), Self-Organizing map (SOM), Support Vector Machine (SVM) and Principal Component Analysis (PCA). SOM was used to determine the level of influence of each rating parameter, K-Means to create groups by clustering, semi-supervised learning with SVM and PCA to generate a model to detect anomalies in the processes. By means of a case study, four groups of processes were obtained, highlighting the presence of the group “null economic offer” where the values for the economic offer do not exceed 1%, and a greater weight is given to other qualification parameters, which include direct contracting. The processes in this cluster are considered anomalous. Following this methodology, a semi-supervised learning model is built for the detection of anomalies, which obtains an accuracy of 95%, allowing the detection of procedures where the aim is to benefit a particular supplier by means of the qualification assignment parameters.

Every day around 3500 people die on the roads and tens of millions suffer injuries or disabilities each year; vehicle manufacturers spend millions of dollars annually on research for the prevention of traffic accidents in development of Advanced Driving Assistance Systems (ADAS). To solve this problem, the development of a model based on decision trees and supervised neural networks is proposed; it is focusing on the four main risk factors of traffic accidents: a) drive inattentive to traffic conditions, b) not respecting the regulatory traffic signals, c) driving in a state of drowsiness or poor physical condition, d) drive vehicle over the Maximum Speed Limits. The Machine Learning system requires three models: a) neural network for object recognition; b) algorithm for sleepiness detection; and, c) decision tree to issue factor-related alerts: driver status (sleepiness or distraction), speed and traffic signal alert. Finally, the model is put into production in a low cost ADAS mobile application, and it was tested in the laboratory using recordings obtained in real driving tests (n = 200, success>=90%) in ideal luminosity environments.