Linear Regression Prediction Model Research Articles

Soft tissue movements after mandibular reconstruction using the vascularized iliac flap: patterns and predictions.

To investigate soft-to-hard tissue response following mandibular reconstruction and to develop a predictive model for projecting soft tissue movement. In this retrospective study, 18 patients receiving mandibular reconstruction using a vascularized iliac flap were enrolled. Various indicators for characterizing the movement of tissues were considered to identify the effective predictors for projecting soft tissue movements. Face-region-specific linear regression models for prediction were constructed and evaluated. The arithmetic mean of hard tissue movement in an extended area had the strongest correlation with the movement of the focal soft tissue, while the arithmetic mean in a regional area (Ram) was a more effective predictor. The linear regression model using Ram, global extrema and distances between them as the predictors performed the best in the lower margin of the face, with an average error of 1.51 ± 1.38mm. Soft tissue movement in the alveolar process was not correlated with the existence of dentition, only can be predicted by the soft tissue movement below it. The area of the masseter was strongly correlation with Ram, but no other factors. An accurate prediction of soft tissue movements in the lower margin and the alveolar process of the face can be achieved by considering hard tissue and adjacent soft tissue movements. No effective predictor in the masseter area was identified. We investigated the relationship between hard tissue movements and the soft tissue responses in the facial area. Through building predictive models for projecting postoperative soft tissue movements, we derive insights for the aesthetic outcome of face surgeries. This study was registered on the Chinese Clinical Trial Registry (registration number: ChiCTR2100054103).

Transport characteristics of heavy metals in the soil-atmosphere-wheat system in farming areas and development of multiple linear regression predictive model

Heavy metal accumulation in agricultural products has become a major concern. Previous studies have focused on the transport of heavy metals from the soil and their accumulation in crops. However, recent studies revealed that wheat leaves, ears, and awns can also transport and accumulate heavy metals. Wheat grains can be influenced by two sources of heavy metals: soil contamination and atmospheric deposition. To comprehend the transport characteristics of heavy metals in soil, atmospheric deposition, and wheat, 37 samples each for wheat rhizosphere soil, wheat roots, stems, leaves, and grains were collected. Fifteen samples of atmospheric dry deposition and atmospheric wet deposition were collected from Linshu County (northern area), China. Based on the test data, the characteristics of heavy metals and their distribution in the study area were analyzed. Migration patterns of heavy metals in crops from different sources were investigated using Pearson correlation and redundancy analysis. Finally, a predictive model for heavy metals in wheat grains was developed using multiple linear regression analysis. Significant disparities in the distribution of heavy metals existed among wheat roots, stems, leaves, and grains. The coefficient of variation of heavy metals in atmospheric deposition was relatively high, indicating discernible spatial patterns influenced by human activities. Notably, a positive correlation was observed between the concentration of heavy metals in wheat grains and atmospheric deposition of Hg, Cd, and Pb. Conversely, Zn and Ni levels in wheat grains were significantly negatively associated with soil Zn, Ni, pH, and OM content. The contribution of heavy metal elements from different sources varied in their impact on the grain's heavy metal content. Specifically, atmospheric deposition was the primary source of Hg and Pb in wheat grains, while Cd, Ni, Cu, and Zn were predominantly derived from soil. Using a multiple linear regression model, we could accurately predict Hg, Pb, Cd, Ni, Zn, and As concentrations in crop grains. This model can facilitate quantitative evaluation of ecological risk of heavy metals accumulation in crops in the study area.

Improving prediction of linear regression models by integrating external information from heterogeneous populations: James-Stein estimators.

We consider the setting where (1) an internal study builds a linear regression model for prediction based on individual-level data, (2) some external studies have fitted similar linear regression models that use only subsets of the covariates and provide coefficient estimates for the reduced models without individual-level data, and (3) there is heterogeneity across these study populations. The goal is to integrate the external model summary information into fitting the internal model to improve prediction accuracy. We adapt the James-Stein shrinkage method to propose estimators that are no worse and are oftentimes better in the prediction mean squared error after information integration, regardless of the degree of study population heterogeneity. We conduct comprehensive simulation studies to investigate the numerical performance of the proposed estimators. We also apply the method to enhance a prediction model for patella bone lead level in terms of blood lead level and other covariates by integrating summary information from published literature.

Recent Advancements and Trends of Topical Drug Delivery Systems in Psoriasis: A Review and Bibliometric Analysis.

Psoriasis is an immune-mediated inflammatory skin disease where topical therapy is crucial. While various dosage forms have enhanced the efficacy of current treatments, their limited permeability and lack of targeted delivery to the dermis and epidermis remain challenges. We reviewed the evolution of topical therapies for psoriasis and conducted a bibliometric analysis from 1993 to 2023 using a predictive linear regression model. This included a comprehensive statistical and visual evaluation of each model's validity, literature profiles, citation patterns, and collaborations, assessing R variance and mean squared error (MSE). Furthermore, we detailed the structural features and penetration pathways of emerging drug delivery systems for topical treatment, such as lipid-based, polymer-based, metallic nanocarriers, and nanocrystals, highlighting their advantages. This systematic overview indicates that future research should focus on developing novel drug delivery systems characterized by enhanced stability, biocompatibility, and drug-carrying capacity.

Research on Automatic Pricing and Replenishment Decision of Vegetable Products Based on Machine Learning

In fresh food supermarkets, due to the short shelf life of vegetable products and the deterioration of products over time, unsold products on the same day cannot be sold the next day. Therefore, supermarkets need to make automatic pricing and replenishment decisions on a daily basis based on historical sales and demand. All product data for this study were provided by the 2023 Higher Education Society Cup National College Student Mathematical Modeling Competition.This study aims to explore the relationship between commodity sales and wholesale prices and pricing using five machine learning methods: neural networks, linear regression, decision tree regression, random forest regression, and XGBOOST regression, in order to predict sales and wholesale prices for the next week. By comparing the performance of five machine learning methods, it was found that the XGBOOST regression prediction model is the most suitable for predicting the daily sales of goods. The linear regression prediction model is the most accurate for predicting the wholesale prices of "chili", "flower and leaf", "edible fungi", "cauliflower", and "eggplant" goods, while the random forest regression prediction model is the most accurate for predicting the wholesale prices of "aquatic rhizome" goods. By constructing a dynamic programming model and using genetic algorithms to solve, automatic pricing and replenishment decisions for various goods within the next week can be obtained.

Prediction of Rock's Brittleness and Dynamic Properties Utilizing Effective Artificial Intelligence Approaches

This research aims to determine the brittleness index (BI) and engineering properties of limestone specimens. In addition, this study evaluates the effect of moisture on the developed models to predict the BI and shear wave velocity (Vs), based on the point load index (Is50), dry and saturated tensile strength (Ts-d and Ts-s), and porosity. Gaussian process regression (GPR), multilayer feed-forward neural network (MFFNN), and multiple linear regression (MLR) predictive models were utilized. Microscopic examination of the limestone specimens revealed that calcite is the predominant mineral. It was observed that samples with higher calcite content exhibited greater brittleness and strength properties while having lower porosity. The results obtained from the MLR analysis demonstrated that it is possible to accurately forecast the brittleness index (BI), as well as the dry and saturated shear wave velocities (Vs-d and Vs-s) at the specific sites under investigation. The moisture effect on developed models showed that Vs prediction in dry conditions (Vs-d) was less accurate compared to the saturated conditions (Vs-s). Conversely, the relationships developed for estimating the BI in dry conditions exhibited higher accuracy. The analysis of all model assumptions using MLR indicated that the models could be reliably utilized. However, the MFFNN and GPR methods were found to be more conservative in estimating these properties. Moreover, the study identified the best transfer function and training algorithm for predicting Vs and BI. The evaluation metrics, such as R2 and RMSE revealed that GPR demonstrated higher precision compared to MFFNN and MLR.

Early Bloomer and Future Master: Getting to the Top of the Modern Badminton World

The road to the top is never easy. This study investigated whether the career trajectories of top-level men’s and women’s badminton players could be predicted by their age at first major debut and the time taken to reach the top 150. Data from the BWF from October 2009 to October 2022 were analyzed using a predictive linear regression model with Bayesian inference, adjusting for anthropometric factors. The results suggested that age at debut influences career high rankings among over 120 elite players. Additionally, observations challenged the Matthew effect in early-career success for young players, proposing that inherent talent may be more significant than early nurturing. The study also examined the potential impact of relative age and early specialization in sports.

Taguchi method and neural network for efficient β‐ketoenamine synthesis in deionized water

AbstractThe optimization of performance parameters, in particular the yield of the synthesis reaction of β‐enaminones in demineralized water, is crucial to improve their efficiency and accuracy. In this report, we investigate the optimization of the synthesis of β‐ketoenamines in deionized water by controlling several parameters such as reaction time, temperature, amine equivalent, acid percentage, and stirring rate. An orthogonal L16 (45) network was created using Taguchi's approach, allowing for the best possible parameters. To forecast the contribution of each parameter, analysis of variance (ANOVA) techniques are also used. Multiple linear and nonlinear regression (MLR, MNLR) and multilayer perception artificial neural network (MLP‐ANN) predictive models were developed. Analysis of the results led to optimized design parameters, with time = 6 h, temperature = 25°C, amine equivalent = 1.5, acid percentage = 20%, and stirring rate = 1000 rpm, leading to a maximum yield of 63%. ANOVA analysis revealed that temperature, stirring rate, acid percentage, and time are the parameters with the greatest influence. The least sensitive parameter is the amine equivalent. The two main interactions are temperature * acid % and amine equivalent * rpm. The MLP‐ANN predictions are in good agreement with the experimental values, resulting in a higher R2 compared to the quadratic regression model and the MLR model. By using molecular docking studies, the produced compounds' biological activity was investigated. Some of the synthesized compounds appear to be interesting and could be used for therapeutic applications. The results of this study give us insight into the gentle, cost‐effective, and biologically active synthesis of β‐enaminones in deionized water.

Artificial neural network and multiple linear regression modeling for predicting thermal transmittance of plain-woven cotton fabric

The present research compares a machine learning model with a statistical model, with specific emphasis on artificial neural networks and multiple linear regression models. The aim of this study is to forecast the thermal transmittance of a plain-woven cotton fabric using input data such as thread density measured in ends per inch, picks per inch, and fabric thickness. The artificial neural network is built using a network with feed-forward backpropagation, and the MATLAB software’s training function trainlm is used to modify its weight and basic values based on Levenberg–Marquardt optimization techniques. The sigmoid transfer function is used to set the layer output and measure network performance in terms of the root mean squared error, mean absolute error percentage, and coefficient of determination which were determined. For the artificial neural network prediction model, the root mean squared error and mean absolute error percentage were 1.05 and 3.132%, respectively, while the coefficient of determination was 0.9307. In contrast, the multiple linear regression prediction model had root mean squared error and mean absolute error percentage values of 2.98 and 8.97%, respectively, along with a coefficient of determination of 0.4727. The results reveal that the artificial neural network model outperforms the multiple linear regression model, showing superior accuracy and robustness in capturing the intricate interactions between important fabric parameters (ends per inch, picks per inch, and thickness) and thermal transmittance values. This research emphasizes the efficiency of artificial neural network modeling as a superior tool for forecasting thermal transmittance in textile applications rather than employing the time-consuming trial-and-error process for delivering significant insights for material engineering and energy-efficient design.

Research on pricing and replenishment strategy of superstore goods based on linear regression and gray prediction models

The pricing and replenishment strategy of various commodities in superstores have a significant impact on the revenue. This paper aims to address the automatic decision-making problem for pricing and replenishment of vegetable commodities in superstores. It applies the linear regression model to predict commodity prices, aiming to maximize the revenue of superstores. Additionally, it utilizes the gray prediction model to forecast replenishment volume. Based on marketing data from 2020 to 2023 for vegetable goods at a specific superstore, this study provides the replenishment quantity and pricing strategy for each vegetable category from July 1-7, 2023, as well as for each individual vegetable item on July 1, 2023.

Predictive modelling of total operating room time for Laparoscopic Cholecystectomy using pre-operatively known indicators to guide accurate surgical scheduling in a critical access hospital

The financial margin of rural and critical access hospitals highly depends on their surgical volume. An efficient operating room is necessary to maximise profit and minimise financial loss. OR utilisation is a crucial OR efficiency metric requiring accurate case duration estimates. The patient's age, ASA, BMI, Mallampati score, previous surgery, the planned surgery, the surgeon, the assistant's level of experience and the severity of the patient's disease are also associated with operative duration. Although complex machine-learning models are accurate in operative prediction, they are not always available in resource-limited hospitals. Laparoscopic cholecystectomy (LC) is one of the most common surgical procedures performed and is one of the few procedures performed at critical access and rural hospitals. The accurate estimation of the operative duration of LC is essential for efficient OR utilisation. We hypothesise that a multivariate linear regression prediction model can be constructed from a set of pre-operatively known, easily collected variables to maximise OR utilisation and improve operative scheduling accuracy for LC. We further hypothesise that this model can be implemented in resource-limited environments, such as critical access hospitals.

Modeling and Detection Scheme for Zero-Dynamics Attack on Wind Power System

In a wind farm, wind power system is key to maintaining continuous generation of point energy. Since Zero-Dynamics attack poses a major risk to future wind farms, investigation on the effect of Zero-Dynamics attack against wind power systems has both theoretical and practical importance. In this paper, we consider a scenario in which a wind power system is subject to Zero-Dynamics attack in power generation chain. Firstly, we transform a semi-direct D-PMSG into an equivalent circuit to establish mathematical model of wind power system under Zero-Dynamics attack. Then, we analyze and prove the advantages and limitations of Zero-Dynamics attack by divide the equation of state into a stable part and an unstable part with the help of system relative degree ρ. In addition, to better achieve desired attack effect, adjustments are made for Zero-Dynamics attack based on its limitations. Finally, we combine an Multiple Linear Regression (MRL) predictive control model with Byrnes-Isidori normal form to protect wind power system. The theoretical results were verified on designed small wind power system.

Research on energy-saving strategies in the architectural design of the Eastern Railway in the era of artificial intelligence

Abstract This paper takes the air conditioning energy consumption in railroad building design as the main object of analysis and uses an orthogonal experimental design method to study the influencing factors and effects of energy consumption in the Middle East railroad construction design. Stochastic working condition simulation is carried out for the typical models of passenger stations in five climate zones to obtain the data set of the passenger station energy consumption model, and the multiple linear regression prediction models of air conditioning energy consumption in different regions are established. Combined with the hour-by-hour average outdoor temperature, hour-by-hour passenger flow and lighting and related equipment data in the waiting hall of a city in July, the air-conditioning cold load of railroad passenger stations is simulated and analyzed. Explore the implementation effect of energy-saving strategies for railroad passenger station air-conditioning system, and for the traditional PID control strategy, put forward the control strategy based on model prediction, analyze the change of centralized air-conditioning system cold load and the change of the actual performance coefficient of the chiller unit under different control strategies. The data show that compared with the traditional PID control, the model prediction control reduces the total energy consumption by 0.2063, which realizes the energy-saving optimization of the air conditioning system for railroad building design.

Semi-parametric single-index predictive regression models with cointegrated regressors

This paper considers the estimation of a semi-parametric single-index regression model that allows for nonlinear predictive relationships. This model is useful for predicting financial asset returns, whose observed behaviour resembles a stationary process, if the multiple nonstationary predictors are cointegrated. The presence of cointegrated regressors imposes a single-index structure in the model, and this structure not only balances the nonstationarity properties of the multiple predictors with the stationarity properties of asset returns but also avoids the curse of dimensionality associated with nonparametric regression function estimation. An orthogonal series expansion is used to approximate the unknown link function for the single-index component. We consider the constrained nonlinear least squares estimator of the single-index (or the cointegrating) parameters and the plug-in estimator of the link function, and derive their asymptotic properties. In an empirical application, we find some evidence of in-sample nonlinear predictability of U.S. stock returns using cointegrated predictors. We also find that the single-index model in general produces better out-of-sample forecasts than both the historical average benchmark and the linear predictive regression model.

Effect of nano-gypsum on mechanical properties cement admixed marginal silty soil

Even though the chemical stabilization is by far the most common ground improvement technique, very few studies have been conducted on the joint incorporation of two or more additives (cement-nanogypsum mix) to stabilise soft soil. This study investigates the effect on the mechanical properties of cement-stabilised silty soft soil (CSS) using nano-gypsum (NG). Unconfined compression strength (UCS) test trials were conducted to achieve soft soil conditions. The effect of NG and cement on soft soil was investigated by performing Atterberg limit, compaction, UCS, and Direct shear tests. The soil was treated with four percentages (10%, 15%, 20%, and 25%) of cement and with three percentages (1%, 1.5%, and 2%) of NG by its dry unit weight. The test results showed that the plasticity index of the soil was reduced significantly by the addition of stabilizers (Cement-NG mix) making the soil stiffer. The compaction tests suggested that optimum cement content was 20% for CSS and for cement-NG mix maximum strength was achieved at cement: NG proportion of 20:1. The unconfined compression strength (UCS) was increased by 9.11 times in comparison to untreated soft soil at cement: NG proportion of 20:1.5. The direct shear test (DST) suggested that there was an increase in cohesion by 4.44 times in comparison to the untreated sample at the soil–cement-NG combination of 78.5: 20: 1.5. However, the angle of internal friction remained almost unchanged. The development of cementitious substances was confirmed through XRD spectroscopy. Microstructural and chemical properties were observed using combined FESEM-EDAX analysis and Fourier transform infrared spectroscopy (FTIR). The multivariate linear regression (MLR) predictive model was developed for the unconfined compressive strength and failure strain as a function of various parameters considered in the study. The test results demonstrated that the combination of cement and NG can result in remarkable improvement in the geotechnical properties of soft soil. The use of NG may also reduce the demand for cement in deep cement mixing technique, which eventually reduces the carbon footprint and renders the ground improvement technique, ecologically beneficial.

Experimental comparison of linear regression and LSTM motion prediction models for MLC-tracking on an MRI-linac.

Magnetic resonance imaging (MRI)-guided radiotherapy with multileaf collimator (MLC)-tracking is a promising technique for intra-fractional motion management, achieving high dose conformality without prolonging treatment times. To improve beam-target alignment, the geometric error due to system latency should be reduced by using temporal prediction. To experimentally compare linear regression (LR) and long-short-term memory (LSTM) motion prediction models for MLC-tracking on an MRI-linac using multiple patient-derived traces with different complexities. Experiments were performed on a prototype 1.0T MRI-linac capable of MLC-tracking. A motion phantom was programmed to move a target in superior-inferior (SI) direction according to eight lung cancer patient respiratory motion traces. Target centroid positions were localized from sagittal 2D cine MRIs acquired at 4Hz using a template matching algorithm. The centroid positions were input to one of four motion prediction models. We used (1) a LSTM network which had been optimized in a previous study on patient data from another cohort (offline LSTM). We also used (2) the same LSTM model as a starting point for continuous re-optimization of its weights during the experiment based on recent motion (offline+online LSTM). Furthermore, we implemented (3) a continuously updated LR model, which was solely based on recent motion (online LR). Finally, we used (4) the last available target centroid without any changes as a baseline (no-predictor). The predictions of the models were used to shift the MLC aperture in real-time. An electronic portal imaging device (EPID) was used to visualize the target and MLC aperture during the experiments. Based on the EPID frames, the root-mean-square error (RMSE) between the target and the MLC aperture positions was used to assess the performance of the different motion predictors. Each combination of motion trace and prediction model was repeated twice to test stability, for a total of 64 experiments. The end-to-end latency of the system was measured to be (389 ± 15)ms and was successfully mitigated by both LR and LSTM models. The offline+online LSTM was found to outperform the other models for all investigated motion traces. It obtained a median RMSE over all traces of (2.8 ± 1.3)mm, compared to the (3.2 ± 1.9)mm of the offline LSTM, the (3.3 ± 1.4)mm of the online LR and the (4.4 ± 2.4)mm when using the no-predictor. According to statistical tests, differences were significant (p-value <0.05) among all models in a pair-wise comparison, but for the offline LSTM and online LR pair. The offline+online LSTM was found to be more reproducible than the offline LSTM and the online LR with a maximum deviation in RMSE between two measurements of 10%. This study represents the first experimental comparison of different prediction models for MRI-guided MLC-tracking using several patient-derived respiratory motion traces. We have shown that among the investigated models, continuously re-optimized LSTM networks are the most promising to account for the end-to-end system latency in MRI-guided radiotherapy withMLC-tracking.

Selection of Agronomic Parameters and Construction of Prediction Models for Oleic Acid Contents in Rapeseed Using Hyperspectral Data

High oleic acid oilseed rape is a hot research area in the development of functional oilseed rape. At present, the model of predicting the oleic acid content in rapeseed at the early growth stage based on hyperspectral technology lacks a mechanistic explanation. In this study, based on the data collected at the 5–6 leaf stage of oilseed rape, a one-dimensional linear regression prediction model of the oleic acid content in leaves (x) and the oleic acid content in rapeseed (y) was constructed with the regression equation y = 1.83x + 75.26, and the R2, RMSE, and RPD of the testing set were 0.96, 0.23%, and 4.86, respectively. Then, a support vector regression prediction model of the spectral standard normal transformed feature parameters and the oleic acid content in leaves was constructed, and the R2, RMSE, and RPD of the testing set were 0.74, 0.21%, and 2.01, respectively. Finally, the sensitive parameter transfer model for the prediction of “spectral standard normal transform feature—oleic acid content in leaves—oleic acid content in rapeseed” was validated, and the R2, RMSE, and RPD of the full sample test were 0.71, 0.54%, and 0.54, respectively. The results show that although the accuracy of the prediction model after the introduction of the agronomic parameters was reduced compared with the performance of direct prediction by using spectra, the oleic acid content in oilseed rape leaves, as an important intermediate variable, could better explain the relationship between the reflection spectrum of the leaf and the oleic acid content in rapeseed. This study provides a theoretical basis and technical support for hyperspectral remote sensing technology in the quality prediction of rapeseed.

Study on the Evaluation Method of Blended Learning Effect Based on Multiple Linear Regression Analysis

With the development of information technology, blended learning has been widely used in the education field, and the evaluation of blended learning effect has become one of the research hotspots. Taking the automobile theory course as an example, a blended learning process with online and offline is designed, and the main learning behaviors that affect learning effect are analyzed. By extracting data on the main learning behaviors of students during the learning process, correlation and linear regression methods are used to analyze the influencing factors of blended learning effect, and a linear regression prediction model is established. The results show that students' online testing, classroom performance, unit testing, feature assessment, and experimental performance are key indicators for predicting learning performance. According to the analysis of influencing factors of blended learning, the countermeasures and suggestions for improving the effect of blended learning are proposed.

Diagenesis and Diagenetic Mineral Control on Reservoir Quality of Tight Sandstones in the Permian He8 Member, Southern Ningwu Basin

The sandstone reservoirs of the He8 member within the Lower Permian Shihezi Formation are important targets for oil and gas exploration in the southern Ningwu Basin. This study utilized thin-section identification, scanning electron microscopy, and X-ray diffraction analysis to examine the petrological features and reservoir characteristics, and evaluate the impact of the mineral composition and diagenesis type on the porosity of the sandstone reservoir. Additionally, a multiple linear regression prediction model was developed to predict the distribution of promising sandstone reservoirs in the study area. The results of the analysis revealed that the sandstone of the He8 member is mainly composed of feldspathic lithic sandstone, followed by lithic sandstone. The main reservoir type is characterized by secondarily dissolved pores and micropores within kaolinite aggregates. The low porosity (ranging from 0.2% to 10.7%) and permeability indicate that the He8 member is a tight sandstone reservoir. This reservoir has undergone compaction, cementation, and dissolution diagenesis, and is presently in the stage of mesodiagenesis B. The rigid framework of quartz, the dissolution of feldspar grains, and the intergranular pores of kaolinite are significant contributors to reservoir quality and the main drivers of porosity. In this study, a multivariate linear regression model was developed based on the mineral content of quartz, feldspar, carbonate minerals, kaolinite, smectite, and rock fragments, which accurately predicts the porosity of the studied reservoirs. Based on this model, it was predicted that the north of the Jingle South sub-depression contains a favorable reservoir space in the tight sandstone reservoir of the He8 member. The findings of this study hold significant reference value in the quantitative characterization of tight sandstone reservoirs with similar depositional and diagenetic characteristics, and improving the prediction effect of favorable reservoirs.

Study on the Drying Process and the Influencing Factors of Desiccation Cracking of Cohesive Soda Saline-Alkali Soil in the Songnen Plain, China

The surface of cohesive soda saline-alkali soil in the Songnen Plain of China exhibits obvious desiccation cracking phenomenon during water evaporation. Quantitative research on the process and influencing factors of desiccation cracking are significant for determining the mechanical properties of salt-affected soil, improving the salinization status and the local ecological environment. This study aims to conduct laboratory-controlled cracking experiments on soda saline-alkali soils, and to regularly measure the soil moisture and the cracking status of 40 soil samples. After processing the crack images uniformly, crack characteristics including crack length, crack area, contrast texture feature, and box-counting fractal dimension are extracted. The results indicate that the decrease in soil moisture can be divided into three stages such as fast evaporation stage, slow evaporation stage and stable evaporation stage, and that the evaporation rate shows a linear negative correlation with soil salinity. The measurements also indicate that crack length rapidly increases and reaches stability with a decrease in water content compared with the crack area. The high correlation between different crack characteristics and the soil total salinity (correlation coefficient from 0.44 to 0.95) shows that salt content is the dominant factor affecting the desiccation cracking of soda saline-alkali soil. A multiple linear regression prediction model was finally established based on crack characteristics for different salinity parameters with high prediction accuracy for Na+, EC and total salinity (R2 &gt; 0.91 and RPD &gt; 2.5) and also certain prediction accuracy for Cl− and CO32− (R2 &gt; 0.66 and RPD &gt; 1.5).